The peak of the spin glass relaxation rate, S(t) = d[ − MT RM (t, tw)/H]/dℓnt, is directly related to the typical value of the free energy barrier which can be explored over experimental time scales. A change in magnetic field H generates an energy Ez = Nsχ f c H 2 by which the barrier heights are reduced, where χ f c is the field cooled susceptibility per spin, and Ns is the number of correlated spins. The shift of the peak of S(t) gives Ez, generating the correlation length, ξ(t, T ), for Cu : M n 6at.% and CdCr1.7In0.3S4. Fits to power law dynamics, ξ(t, T ) ∝ t α(T ) and activated dynamics ξ(t, T ) ∝ (ℓnt) 1/ψ compare well with simulation fits, but possess too small a prefactor for activated dynamics.PACS numbers: 75.50. Lk, 75.10.Nr, 75.40.Gb The study of the irreversible behavior of the spin glass magnetization under a change of magnetic field allows exploration of the available states of a random frustrated system. [1,2] There are various representations for the long time evolution and the dynamics of spin glasses, [3-5] but a coherent, overall accepted real space description remains lacking.[6] The purpose of this paper is to extract a time and temperature dependent spin glass correlation length from a specially structured set of experiments, and to compare our results with available theoretical predictions.The definition of a correlation length for a spin glass is difficult to express in measurable terms. Marinari et al. [7] and Kisker et al. [8] introduced the time dependent equal time correlation function at time t. In the notation of Ref. 7,where the average is done at time t, and σ i (σ i+x ) and τ i (τ i+x ) represent the z component of Ising spins at sites i (i + x) in two thermalized configurations in a box of volume V . To avoid accidental contributions to G(x, t), the two configurations are chosen to have zero overlap, q = V −1 i σ i τ i . Refs. 7 and 8 both observed, through their simulation studies, that for large times t the correlation function G(x, t) differs from zero for distances not too much larger than a dynamic correlation length ξ(t, T ). Simulations of Marinari et al. [7] obtain satisfactory fits for ξ(t, T ) ∝ (t/τ 0 ) αT /Tg , appropriate to power law dynamics, [4] while Kisker et al. [8] fit satisfactorily both this proportionality and equally well ξ(t, T ) ∝ [(T /T g )ℓn(t/τ 0 )] 1/ψ , appropriate to activated dynamics. [5] Our measurements consist of cooling a sample in a magnetic field through the glass temperature T g to the measuring temperature T , waiting a time t w , then cutting the field to zero and measuring the decay of the magnetization. This generates the response function,where M T RM (t, t w ) is the thermoremanent magnetization at time t after cutting the magnetic field to zero. Our approach, justified previously through magnetic field cycling [9] and used to determine the Parisi physical order parameter P (q), [1] makes use of the scaling relationship introduced by Vincent et al. [2]. They show that barrier heights surmounted during aging are reduced upon a...