Picture a trade caravan setting out. It has a destination — somewhere with water, wealth, rest — and it never gets there.
Not because it's ambushed once and turned back. Because the road itself is built out of things that look like arrival and aren't.
There's a shimmer on the horizon that reads as a city, full and lit and populated. The caravan detours toward it and it isn't there. There's a dry riverbed that, in the heat, looks exactly like a river, and someone runs for it with a water skin.
There's a stand of trees offering shade that turn out to be poisonous. Six bandits work this road continuously, and they know exactly when to strike: the moment you're distracted by the mirage.
This is how the Bhāgavata Purāṇa describes ordinary life. Not once — twice, back to back, in successive chapters, at two different lengths, as if the first telling wasn't going to be believed and needed saying again in more detail.
The short version comes first, compressed into a page. The caravan, the six bandits, a python that swallows travelers whole, a stand of poisonous trees, and this closing image:
"Sometimes, a city of the gandharvas is seen. Here and there, evil spirits fleetingly appear, like meteors."1
A gandharva-city is the Sanskrit stock image for a mirage — a shimmering settlement that exists only as heat distortion, dissolving the moment you approach it. It shows up in this text and in a dozen adjacent ones as shorthand for the goal that looks real from a distance and stops existing up close.
Both tellings use it as the anchor image, and both circle back to it more than once — a sign, in a text otherwise built from lineage lists and ritual instruction, that this particular picture was doing more work than the surrounding prose.
Then, one chapter later, the same allegory returns — longer, slower, with new detail layered in, including a passage that repeats itself inside its own retelling:
"Looking at each other's faces, they indulge in carnal acts and forget that time has a limited span."2
That sentence appears twice within the second telling alone, nearly word for word, describing two different scenes. The text doesn't flag the repetition either time — not chapter to chapter, not within the chapter. It just says the same true thing again, as if repetition were part of how the point gets made rather than an editing slip.
The first pass gives you the shape: caravan, bandits, mirage, danger. The second pass gives you the texture of getting closer, which is where the allegory earns its length.
A man chasing wealth is compared to someone hunting fire and finding instead an ulmuka-pishacha — literally a fire-demon, the will-o'-the-wisp phenomenon of marsh gas igniting in the dark, which looks exactly like the fire you wanted until you reach it and there's nothing there.3
Gold itself gets renamed mid-sentence as "a special kind of excrement,"4 which is not a moral aside. It's the allegory's whole method: restating a valuable as its own opposite so the desire for it looks strange from outside, for one second, before the sentence moves on.
And it adds a detail the short version skips entirely: the caravan sometimes succeeds. People do reach a gandharva-city and, for a while, live in it. "Some enter a city of the gandharvas and are delighted for a while, like someone who has attained success."5
That clause — for a while — is doing the real work of the whole passage. This isn't a story about a trick that always fails immediately. It's a story about a trick that sometimes works long enough to be believed in.
The single strangest image in either telling isn't a bandit or a mirage. It's this:
"Monetary transactions with each other increase the ties of enmity with each other."6
Read that against the frame it sits in. These are travelers on the same caravan, headed toward the same destination, facing the same six bandits — and the text's account of their downfall isn't that outsiders rob them. It's that they rob each other, through ordinary commerce, and the commerce itself is what generates the hostility.
Trading with a fellow traveler should be the safest transaction on the whole road. It's the one the text singles out as corrosive.
The mechanism, once you sit with it, is close to something modern behavioral economics has a name for. A merchant who has already sunk cost into the caravan — time, goods, the relationships built along the way — keeps transacting inside it even as the transactions erode the thing they were meant to build, because stopping would mean admitting the investment already made hasn't paid off.
The caravan doesn't dissolve from an external shock. It corrodes from continued participation in a system everyone privately suspects isn't working, because each person's stake in having already participated is what keeps them participating. Nobody outside the caravan needed to attack it. The travelers did the work themselves, one honest-looking transaction at a time.
It would be easy to file the doubled telling as sloppy editing — Debroy's endnotes flag genuine contradictions elsewhere in this translation and this repetition earns no note at all, which itself is informative.7 But look at what repeats. It isn't a stray detail. It's the single most concrete, most human line in the entire allegory — people looking at each other's faces, forgetting time is short — landed twice, unremarked, inside a passage that is itself the second of two full retellings of the same journey.
That's a structure, not a slip. The forest of saṃsāra is a place the text insists on describing more than once because the thing it's describing is a place you go through more than once. Nobody hears the caravan allegory a single time and stops being susceptible to gandharva-cities. The text's own form — telling it twice, and inside the second telling, repeating its sharpest line — enacts the very point it's making: recognizing the mirage doesn't inoculate you against the next one.
Two more hazards on this road, easy to skip past in a list this long, and each one earns its place by naming something the mirage and the bandits don't quite cover.
The first is sleep, and the text doesn't soften the image: "Sometimes, the python of sleep devours him. Immersed in that blinding darkness, he lies down, as if in a desolate forest. Like a dead body that has been flung away, he does not know anything."10 Not rest. Not recovery. A predator that swallows you whole and leaves you indistinguishable from a corpse for the duration. The caravan doesn't merely risk ambush while awake — its travelers periodically vanish from their own journey altogether, unconscious, for stretches they don't remember and can't account for afterward.
The second is smaller and stranger: a man searching for "drops of honey to satisfy his desire," who gets stung for the trouble and, even when he does manage to extract something, has it "forcibly robbed by someone else."11 The honey isn't a mirage — it's real, it exists, a small true pleasure sitting inside something that will hurt you to reach and be taken from you the moment you have it. That's a different failure mode from chasing a gandharva-city. The city was never there. The honey is there, and having it doesn't help.
Between them, sleep and honey cover the two things the mirage-and-bandits framing leaves out: the traveler who loses time to unconsciousness rather than distraction, and the traveler who gets exactly what he wanted and finds the getting was the whole problem.
Neither hazard requires a mistake. You don't have to be fooled by a mirage to fall to either one — sleep comes for the vigilant and the careless alike, and the honey is genuinely sweet. That's what makes them the allegory's two hardest cases: everything else on this road can be avoided by better judgment. These two can't.
You start a project with a clear destination in mind — the finished thing, the launch, whatever counts as arrival. Early on, the road is honest: hard work, visible progress, real distance covered.
Then, somewhere in the middle, something appears that looks exactly like the destination. A version of the thing that's almost done, that could plausibly ship, that a few people say is impressive. You slow down and orient toward it, the way the caravan orients toward the shimmer on the horizon.
Get close and it isn't finished. It was never finished — it was a stage of the work that, from a middle distance and under enough fatigue, reads as an ending. This isn't a failure of judgment so much as a structural feature of the road: the closer you are to genuinely tired, the more anything shaped like relief reads as arrival.
Here's the part the allegory would want you to notice, because it's the part that actually recurs: you do this more than once on the same project. The second gandharva-city doesn't look like the first one. It has learned from your last disappointment and disguises itself as something more specific — not "done," but "good enough for now," or "what the market actually wants," or some other reframing that arrives exactly when you're most tired of not being finished. Recognizing the first mirage buys you nothing against the second, because the second one is built, in effect, out of the lesson you learned from the first.
The caravan doesn't fail because nobody warned it. It fails because the warning wears out and the road keeps generating new shapes for the same trick to wear.
Named once, briefly, and easy to lose in the longer allegory: six bandits work this road, and the text doesn't wait until later chapters to specify who they are — they're the six enemies named elsewhere in this same volume as desire, anger, greed, delusion, pride, and envy.8 What the forest allegory adds to that list is timing, not membership: "They are like jackals and wolves who enter when one is distracted and carry away lambs."9
They don't attack at the start of the journey, when attention is sharpest, or during the featureless stretches, when there's nothing to be distracted by. They attack exactly when a mirage has pulled focus — which means the mirages and the bandits aren't two separate hazards on this road. The mirage is the bandits' opening. Every gandharva-city is also, simultaneously, the moment the six enemies get their clearest shot, and the text's insistence on describing the mirages in such loving, specific detail is really an insistence on describing exactly when you're least defended.
Strongest support: the doubled telling is not an isolated oddity in this text. The translator's own endnotes flag other unflagged repetitions and contradictions across this volume, which means "the text repeats itself without saying so" is a documented pattern of the translation as a whole, not a one-off feature of this passage. [SCHOLARLY — translator's introduction and endnotes]
Tension. The allegory's implicit advice — recognize the mirage, don't chase it — sits awkwardly beside the text's insistence, elsewhere in this same volume, that engaged worldly action performed without attachment is a legitimate path. The forest allegory reads as a wholesale indictment of pursuit; the surrounding chapters carve out an exception for pursuit-done-correctly. The text does not reconcile these, and this page does not resolve it on the text's behalf.
Open question. Debroy's endnotes catch a real number of contradictions and inconsistencies across this volume and flag them explicitly. This repetition gets no flag at all. Is that because the translator read it as intentional repetition-for-emphasis rather than a seam, or because it simply didn't register against the density of everything else in these chapters? Not answered here — worth checking against Debroy's other volumes for whether he ever comments on this specific device.
Compare this to how the Śiva Purāṇa material already in this vault handles repeated material: that text has at least four unflagged doublets across its volumes — the same episode told twice with incompatible details, and the translator's endnotes catch none of them. The forest of saṃsāra is a different case entirely. Here the content of the two tellings agrees completely; only the length and the specific illustrative detail change. It isn't a doublet in the sense of conflicting accounts. It's closer to a chorus — the same claim, sung twice, at different volumes. Worth distinguishing these as two separate translator-layer phenomena in this vault's growing catalogue of how these texts repeat themselves: conflicting doublets (Śiva Purāṇa) versus reinforcing repetition (this passage), and worth checking whether Debroy's endnote behavior differs systematically between the two.
With Sunk Cost Fallacy — Why the Past Haunts Forward-Looking Decisions (psychology). That page's account of why the fallacy persists is unusually charitable to it: sunk cost reasoning isn't stupidity, it's "a sophisticated adaptation to a world where commitment matters" — without it, people would abandon every uncertain venture at the first sign of difficulty, and nothing long-term would ever get finished. The fallacy is adaptive in some contexts and maladaptive in others, and the page notes humans haven't developed the ability to toggle cleanly between them.
The caravan's internal trading — travelers transacting with each other in ways that increase enmity rather than resolve it — is that same toggle failure, dramatized. Every trade inside the caravan is, structurally, a decision about whether to keep honoring a commitment (to the group, to the shared journey, to the goods already carried this far) or to cut losses. The allegory shows the mechanism running in the maladaptive direction continuously: commitment to the caravan becomes the thing that generates the caravan's corrosion, because leaving would mean admitting the investment in it hasn't paid off. What neither page states alone: the psychology page treats sunk cost as a decision bias that operates on an individual choosing whether to persist. The allegory shows what it looks like distributed across a whole group who each privately suspect the shared venture is failing and each keep transacting inside it anyway — not from individual irrationality but because everyone else's continued participation is the strongest evidence available that participation is still rational. Sunk cost, run socially, becomes self-reinforcing in a way the individual-choice framing doesn't fully capture.
Sharpest implication. The text tells this allegory twice and repeats its sharpest line inside the second telling, and that structure is the argument: recognizing one mirage does not make you immune to the next, because the next one is built, in part, out of what you learned from the last. Whatever finally convinced you the last gandharva-city wasn't real is exactly the material the next one will be built from.
Generative questions.