A binary branching process is considered with time-dependent birth and death rates \(\lambda_ t\), \(\mu_ t\), \(t\geq 0\), which are constant on intervals \([n,n+1)\), and \((\lambda_ n,\mu_ n)\), \(n=0,1,2,...\), is a sequence of i.i.d. pairs (“random environment”). In the critical case \(E(\lambda -\mu)=0\), a limit theorem is proved for \(t^{-1/2} \ln \xi_ t\), where \(\xi_ t\) is the number of particles at time t, under the condition of non-extinction \(\xi_ t>0\). Thus, the process in random environment behaves different from the classical process with constant birth and death rates.