[an error occurred while processing this directive]
On this page:

How long is a generation?

This page describes results of an analysis of generation lengths in trees submitted to the project. We do not claim this as a representative sample, but results do agree with others' findings and may shed a bit more light than general statements.


The time unit that DNA respects is transmission events or generations, in which DNA is passed from parent to child. Yet, DNA matches often relate to the number of generations of separation between two living people descended from a common ancestor. Or, they may be described in terms of cousinship -- 2nd, 3rd, 4th, etc.

Neither generations of separation or cousinship, of course, directly say when that ancestor may have lived. Genealogists, and especially genetic genealogists, have tried to estimate years per generation as a means of estimating those dates. There is much controversy about which estimate is best.


Taylor Family Genes reviewed 325 lineage trees submitted to the project for publication. Of those, 245 (75%) contained sufficient data to enable analysis.

We used the birth year of the earliest known (most distant) ancestor (abbreviated EKA) and number of generations from him to the submitting descendant, along with the submittal date of the tree. Submittal date minus EKA birth yielded the number of years included in the tree and the result was divided by number of generations to get an average generation length in the tree.

This is the "Raw" number. The "Adjusted" number assumes the submitter was born 55 years before submittal; we did not collect the birth dates of the submitters. In effect, the adjustment subtracts 55 years from the most recent generation. .

Here are summary statistics from that analysis:

n = 267 Average Median Mode St. Devn. Min Max
No. Generations 6.31 6.00 6.00 2.92 2 22
EKA Birth Year 1764 1791 1800 91.22 1256 1947
Years/Gen 39.14 37.86 40.00 7.70 26.50 90.00
29.27 29.00 29.00 5.45 17.67 62.50

The above statistics suggest that the "Raw" figures are too high and the "Adjusted" figures too low. A reasonable compromise seems somewhere between the two.


Notice the wide disparity from one family to another as displayed in this histogram

Focusing on the 20-25 year range:

Notes on the Data

Although taking some precautions to limit errors, we can not vouch for accuracy of the submitted data. For example, one tree with an EKA birth of 1610 listed only four generations since -- for an average generation length of 90 years! We believe at least five generations were omitted and a more credible average generation length in this family is 44 years "raw", 38 "adjusted".

We did not calculate generation lengths within paternal genetic families.


We draw these conclusions from the data:

  1. Generation length is highly variable from one family to another. 
  2. A "most probable" estimate has low confidence of being correct. It has, at best, only 22% probability of being right within ±1 year and 54% within  ±3 years.
  3. Confidence ranges are large. A 90% confidence interval ranges from 22 to 49 years.