The lack of specificity of diffusion tensor MRI (DT-MRI) arises mainly from its inadequacy to characterise more than one fibre population within a voxel and its implicit assumption of there being a single tissue compartment. This has motivated the search for more sophisticated approaches to modeling diffusion in complex tissue. Hybrid models like CHARMED [1] and NODDI [2], expressing the signal as a summation over intra- and extra-axonal compartments, have proven particularly useful in this context. The intra-axonal volume fraction is often considered a proxy for the axonal density. These frameworks have been extended to characterize axonal diameters [3,4], providing the opportunity to non invasively map the distribution of axonal calibres within the brain, increasing the specificity of diffusion MRI even further. When measuring the signal coming from different water pools, one assumes the T2 relaxation time of every pool to be the same; if that’s not the case, the measured compartment fraction will be weighted by the corresponding T2, and experiments performed with different echo times (TE) would return different values for the axonal density (and, as a consequence, bias the measures of the axonal diameter). Normally, TE is chosen automatically as the minimum allowed by the sequence and is thus very likely to be different from one scanner to the other. Here, we investigate the influence of the echo time on the intra-axonal volume fraction as measured with the CHARMED model. 1 rat underwent a diffusion MRI protocol at 9.4T comprising a diffusion-weighted spin echo with b-value 1000 and 3000 s/mm2, 15 directions in each shell and 3 b=0 scans. The protocol was run with the TE minimisation on, which gave TE=30ms and a single average. Then, a second protocol was acquired with TE=50ms and 2 averages. Assuming an average T2 relaxation of 45ms [5], the increased number of averages compensates for the increased attenuation to have the same SNR in the two scans. Data were analysed in native space using custom software written in Matlab R2012b (The Mathworks) using conventional DT-MRI and the CHARMED model, to generate maps of the fractional anisotropy (FA) and of the intra-axonal volume fraction (FR), respectively. Then, the FA maps were nonlinearly warped to a rat template in stereotaxic coordinates [6] using FSL [7] and the same transformation was applied to the FR maps. Mean values and associated standard deviations were calculated for 31 regions of interest (ROI) segmented on the template; then, for each ROI, the difference between FA/FR calculated at the shortest TE and the same parameters calculated at the longest TE was calculated.While FA does not show any particular trend for increasing TE, FR shows a clear trend where all ROIs except 2 have largest FR for the largest TE. Paired t-test confirms that FR mean values across all ROIs calculated at different echo times are statistically different (p<0.01), while FA values are not. The mean FR difference across all ROIs is -12%. This is compatible with the intra-axonal water being proportionally less attenuated for increasing TEs, i.e., a longer T2 for the intra-axonal water. In addition, the difference between intra- and extra-axonal T2 relaxation times is compatible with results obtained with other methods [8].The obtained results indicate that 1) caution is needed when comparing results for the restricted water fractions obtained using different sequences, or acquired in different centres; and 2) T2 relaxation properties of intra- and extra-axonal water pools, believed to be so similar to be indistinguishable, can be instead accessed using diffusion. This can give access to a new kind of contrast that has not been explored before: by collecting more datapoints, the full T2 decay can be measured for the two pools. In areas of crossing fibres, there is also potential to combine T2 relaxation with diffusion, as already done for T1 relaxation [9], to get fibre-specific values of T2. The acquisition of more samples and more TEs is in progress.