For the first 100 years of modern aphasia research (~since the 1860s) the focus was mainly on syndromes: motor (Broca's) aphasia, sensory (Wernicke's) aphasia, conduction aphasia, and so on. Things changed after the Cognitive... er, make that the Information Processing Revolution (see discussion here or here) when symptoms came more into focus. The symptom approach was an important advance and is still dominant, as the popular method, voxel-based lesion-symptom mapping (VLSM), highlights.
But mapping symptoms isn't the goal. What we are really after are the computations that underlie the symptoms. Recent work by Gary Dell and colleagues suggests that this might be possible using what they call, voxel-based lesion parameter mapping (VLPM). The basic idea is this. Start with an explicit computational model of the process of interest; Dell et al. use the two-step naming model. Collect data on the symptoms in brain injured patients; for Dell et al, this is the distribution of types of naming errors. Adjust the parameters of your model to fit each patient's error pattern; semantic and phonological weights in the Dell model. Then run your voxel-based analysis using the parameter values as your dependent measure instead of say error rate. And it actually works. To the extent that the model is a reasonable approximation to what's going on in the wetware (no model is "right" of course), you are now mapping computations. Pretty cool. All we need now is to improve our models.
Dell, G. S., Schwartz, M. F., Nozari, N., Faseyitan, O., & Branch Coslett, H. (2013). Voxel-based lesion-parameter mapping: Identifying the neural correlates of a computational model of word production. Cognition, 128(3), 380-396. doi: 10.1016/j.cognition.2013.05.007