The Mathieu group M 23 of order 2 7 Á 3 2 Á 5 Á 7 Á 11 Á 23 is one of the 26 sporadic groups. In this note we determine the cohomology rings H Ã M 23 Y F p for all primes p dividing jM 23 j. This has several interesting consequences.Among the sporadic groups M 23 is somewhat unusual in that OutM 23 MultM 23 1Yand, as a result of our calculation H i M 23 Y Z 0 for i`5. In particular M 23 is the ®rst known counter-example to the conjecture that if G is a ®nite group with H i GY Z 0, i 1Y 2Y 3, then G f1g. This conjecture was orginally made by Loday in the mid-1970s and it is based on Quillen's calculations of the cohomology of many of the families of groups of Lie type a few years earlier; see Gi¨en [10]. The ®rst Mathieu group M 11 and the ®rst Janko group J 1 also satisfy OutG MultG 1, but for both of these groups H 3 GY Z H 0. It would be tempting to amend the conjecture. It is very likely that it only fails for a very small number of the sporadic groups among the simple groups. So one might well suspect that there is a (small) ®nite number n so that H i GY Z 0 for 0`i n implies that G f1g if G is ®nite. But we have no idea as to a suitable candidate for n. At the prime 2 we have that H 6 M 23 Y Z Za2 is the ®rst non-zero homology group. In particular, when we look at the usual inclusion M 23 r S 23 we can ask about the image of this ®rst non-trivial class. We have the well known equivalence B S y p Q 0 S 0 , the 0-component of the in®nite loop space lim n3y n S n , and at the prime 2, Q 0 S 0 2 p imJ 2 Â CokerJ 2 is the Quillen±Sullivan splitting; see [13]. Then CokerJ 2 is 5-connected and we have Theorem. The composition B M 23 3 B S 23 3 B S y Q 0 S 0