We thank Drs Maddock and Buonocore (2012) for their comments and interesting arguments in relation to our article (de la Fuente-Sandoval et al, 2011). In their letter, Maddock and Buonocore raised a concern about our report. They suggested that limitations in the quantitative methods used in this paper appear to invalidate the conclusions. There are some reasons, however, to argue against this possibility.

First, all spectra were water-scaled and processed identically. Water scaling is automatically done by using the LCModel with a GE single-voxel phased array (LCModel manual, version 6.2-3, p 122). Our data were acquired in a 3-T GE scanner (Signa Excite HDxt) with a phase-array antenna and a high-resolution eight-channel head coil. We apologize that this detail in the analysis was not explicitly stated in the original manuscript.

Second, systematic differences in MRI and subject's head volume and position would affect not only glutamate or N-acetyl-aspartate, but also likely all the measured metabolites. In addition, these MRI system and subject factors would affect both studied voxels (dorsal caudate and cerebellum) rather than just one brain region (dorsal caudate for glutamate).

Finally, although creatine normalization is a commonly employed method in 1H-MRS studies, we avoided it, as some studies have reported altered levels of creatine in both prodromal and first-episode patients (Bustillo et al, 2002, 2008; Wood et al, 2003; Stone et al, 2009).

Regardless of these arguments, we do acknowledge the inherent limitations of this imaging technique (Lahti and Reid, 2011). Longitudinal studies with larger samples are needed to determine whether glutamate alterations can truly predict conversion to psychosis.