We study the order statistics of one dimensional branching Brownian motion in which particles either diffuse (with diffusion constant D), die (with rate d) or split into two particles (with rate b). At the critical point b = d which we focus on, we show that, at large time t, the particles are collectively bunched together. We find indeed that there are two length scales in the system: (i) the diffusive length scale ∼ √ Dt which controls the collective fluctuations of the whole bunch and (ii) the length scale of the gap between the bunched particles ∼ D/b. We compute the probability distribution function P (g k , t|n) of the kth gap g k = x k −x k+1 between the kth and (k +1)th particles given that the system contains exactly n > k particles at time t. We show that at large t, it converges to a stationary distribution The statistics of the global maximum of a set of random variables finds applications in several fields including physics, engineering, finance and geology [1] and the study of such extreme value statistics (EVS) has been growing in prominence in recent years [2][3][4][5][6][7]. In many real world examples where EVS is important, the maximum is not independent of the rest of the set and there are strong correlations between near-extreme values. Examples can be found in meteorology where extreme temperatures are usually part of a heat or cold wave [8] and in earthquakes and financial crashes where extreme fluctuations are accompanied by foreshocks and aftershocks [9][10][11][12]. Nearextreme statistics also play a vital role in the physics of disordered systems where energy levels near the ground state become important at low but finite temperature [4]. In this context, the distribution of the kth maximum x k of an ordered set {x 1 > x 2 > x 3 ...} (order statistics [13]) and the gap between successive maxima g k = x k − x k+1 provides valuable information about the statistics near the extreme value. Such near-extreme distributions have recently been of interest in statistics [14] and physics [15,[17][18][19]. Although the order and gap statistics of independent identically distributed (i.i.d.) variables are fully understood [13], very few exact analytical results exist for strongly correlated random variables. In this context, random walks and Brownian motion offer a fertile arena where near-extreme distributions for correlated variables can be computed analytically [16][17][18][19].Another interesting system where order statistics plays an important role is the branching Brownian motion (BBM). In BBM, a single particle starts initially at the origin. Subsequently, in a small time interval dt, the particle splits into two independent offsprings with probability b dt, dies with probability d dt and with the remaining probability (1 − (b + d) dt) it diffuses with diffusion constant D. A typical realization of this process is shown in Fig. 1. BBM is a prototypical model of evolution, but has also been extensively used as a simple model for reaction-diffusion systems, disordered systems, nuclear reactio...