Culture · Essay
Pitru Paksha: sixteen days owed to the last three generations
For a fortnight each autumn Hindu families feed their dead by name, three generations at a time, and hand-written priestly ledgers at Haridwar and Gaya record who came. On shraddha, pinda daan, the crows, and India's living family archive.Written to last.
By Confinity Editorial · 2026-08-21 · 6-minute readQuiet tools, kept out of the way.
There is a moment in the shraddha rite when the priest asks for names. Not a general blessing on the family dead: the actual names, of the father, the grandfather and the great-grandfather, in order, with their gotra, and then the same on the mother's side. The offering is made to each of them separately. If a name has been lost, that is a problem, and the family will usually have to say so out loud in front of everyone present.
Pitru Paksha is a fortnight built on that requirement. For sixteen lunar days each autumn, Hindu households across India and Nepal suspend anything auspicious, and turn instead to the pitrs, the ancestors, who are understood to be near and to be owed something specific. It is one of the few remembrance traditions anywhere that will not function on sentiment alone. It needs the record.
Two things are done, and they are usually done together.
Tarpan at the river in Murshidabad. The water is poured through a particular part of the hand, and the names are spoken as it falls. Photograph by Kiran Guha, CC BY-SA 4.0, via Wikimedia Commons.
Tarpana is the offering of water. Standing in a river or beside a vessel, the performer pours water through the hand with black sesame in it, letting it fall from a particular part of the palm, and speaks the names as he pours. Water is offered first to the gods, then to the sages, then to the ancestors, and the direction and the finger used differ for each.
Pinda daan is the offering of food. Balls of cooked rice, sometimes barley flour, worked together with ghee, black sesame and honey, are shaped by hand and set out on a bed of kusha grass. One pinda is made for each of the three ascending generations. The intent is not symbolic feeding in a vague sense: the belief is that the dead have a body of a particular kind that these offerings sustain, and that a descendant is the only one who can do it for them.
The crows come into it at the end. The food is set out for them, and their arrival is taken as the ancestors accepting what was offered. If no crow comes, families notice, and in some regions the offering is repeated. Crows in this tradition are messengers with a right of first refusal.
The whole fortnight closes on Sarvapitri Amavasya, the new moon, which acts as a catch-all: it covers everyone whose date of death is not known, everyone who died in circumstances that left no anniversary, and anyone the family has failed to name during the preceding fifteen days. A calendar that plans for its own gaps is doing something quite sophisticated.
Certain places are held to make the rite more efficacious, and the first among them is Gaya in Bihar, where offerings are made along the Falgu river and at the Vishnupad temple. During Pitru Paksha the town holds a pitrapaksha mela and receives hundreds of thousands of people who have travelled to make an offering their home altar could also have received, on the understanding that this ground carries it further. Varanasi, Haridwar, Prayagraj, Nashik and Rameswaram serve the same function.
An entire hereditary profession exists around this. The pandas, priests attached to particular pilgrimage towns, conduct the rites for visiting families, and each panda serves particular regions, castes and clans by inheritance rather than by arrangement.
Here is the part of Pitru Paksha that has no real equivalent anywhere else in the world, and that almost nobody outside India knows about.
The pandas keep books. When a family comes to make an offering, the presiding priest writes the visit into a hand-bound register: the names of the dead being honoured, the names of those present, the village, the caste and clan, the date, and frequently a signature from the head of the family. The next time anyone from that family comes, in five years or in fifty, the panda finds the earlier entry and adds to it.
The result, at Haridwar in particular but also at Gaya, Kurukshetra, Varanasi, Trimbakeshwar and Chintpurni, is a distributed genealogical archive of Hindu families maintained continuously by hand for centuries. Some registers are said to run back several hundred years. They are held privately, organised by region and clan rather than by surname, and they are consulted: people travel to Haridwar specifically to have a panda pull the family volume and read out who came before them.
It is a striking thing to sit with. A religious obligation to name your ancestors correctly produced, as a side effect, one of the largest and longest-running family record systems on earth, kept in ink, in private hands, without a state, a database or an institution behind it. The record exists because the rite could not be performed without it.
The fortnight ends where another festival begins. In Bengal, the new moon that closes Pitru Paksha is Mahalaya, and it marks the turn from the ancestors to the goddess: the beginning of Devi Paksha and the countdown to Durga Puja.
Since 1931, Mahalaya morning in Bengal has begun before dawn with a radio broadcast, Mahishasura Mardini, a recitation from the Chandi with songs, associated above all with the voice of Birendra Krishna Bhadra. Households wake at four in the morning to hear it. An attempt to replace him in 1976 with a film star reading the same script was so badly received that the original recording was restored within weeks and has been broadcast ever since.
So the fortnight of the dead ends, in one part of India, with several million people getting up in the dark to listen to a man who died in 1991. Which is its own form of ancestor observance.
Traditionally the rite falls to a son, and preferably the eldest, which has consequences that are not incidental. Sons have been valued in part because they are the ones who can perform shraddha. A family without a son has historically faced a real problem about who will feed the dead.
That has been shifting, unevenly and with argument. Daughters performing pinda daan and lighting funeral pyres is no longer rare, has been publicly supported by some priestly bodies, and remains contested in others. The scriptural position is more open than the customary one: several texts allow for a daughter, a grandson through a daughter, or another relative in the absence of a son.
Almost every other tradition of the dead can be performed with feeling alone. You can grieve without a name. You can lay flowers without a date. Pitru Paksha cannot be done that way. It requires you to know exactly who your great-grandfather was, and to say so, and if you do not know, the rite exposes it in front of the whole family.
That is an uncomfortable design and a very effective one. It converts remembering from a private good intention into an annual public test, with a written record kept by someone else and consulted the next time round.
It is a test most families elsewhere would fail and would rather not sit. The pandas at Haridwar keep their registers precisely because somebody worked out, a very long time ago, that four generations is further than memory reaches on its own. The rite did not trust anyone to simply remember. It asked for the writing first, and then made the reading annual.
What the rite actually consists of