We have measured the 2-particle correlation function of atoms from a Bose-Einstein condensate participating in a superradiance process, which directly reflects the 2nd order coherence of the emitted light. We compare this correlation function with that of atoms undergoing stimulated emission. Whereas the stimulated process produces correlations resembling those of a coherent state, we find that superradiance, even in the presence of strong gain, shows a correlation function close to that of a thermal state, just as for ordinary spontaneous emission.PACS numbers: 03.75. Kk, 67.10.Jn, 42.50.Lc Ever since the publication of Dicke's 1954 paper [1], the problem of the collective emission of radiation has occupied many researchers in the fields of light scattering, lasers and quantum optics. Collective emission is characterized by a rate of emission which is strongly modified compared to that of the individual atoms [2]. It occurs in many different contexts: hot gases, cold gases, solids and even planetary and astrophysical environments [3]. The case of an enhanced rate of emission, originally dubbed superradiance, is closely connected to stimulated emission and gain, and as such resembles laser emission [4]. Lasers are typically characterized by high phase coherence but also by a stable intensity, corresponding to a Poissonian noise, or a flat 2nd order correlation function [5]. Here we present measurements showing that the coherence properties of superradiance, when it occurs in an ultracold gas and despite strong amplified emission, are much closer to those of a thermal state, with super-Poissonian intensity noise.Research has shown that the details of collective emission depend on many parameters such as pumping configuration, dephasing and relaxation processes, sample geometry, the presence of a cavity, etc. and, as a result, a complex nomenclature has evolved including the terms superradiance, superfluorescence, amplified spontaneous emission, mirrorless lasing, and random lasing [2,4,[6][7][8][9], the distinctions among which we will not attempt to summarize here. The problem has recently seen renewed interest in the field of cold atoms [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25]. This is partly because cold atoms provide a reproducible, easily characterized ensemble in which Doppler broadening effects are small and relaxation is generally limited to spontaneous emission. Most cold atom experiments differ in an important way from the archetypal situation first envisioned by Dicke: instead of creating an ensemble of excited atoms at a well defined time and then allowing this ensemble to evolve freely, the sample is typically pumped during a period long compared to the relaxation time and emission lasts essentially only as long as the pumping. The authors of reference [10] however, have argued that there is a close analogy to the Dicke problem, and we will follow them in designating this process as superradiance.In the literature on superradiance there has been relatively little discussion...