As I mentioned previously, one problem with the claim the semantic dementia involves an amodal semantic deficit is that the vast majority of the tests use visual, auditory, and language related measures, all of which rely on temporal lobe systems. And as Johnathan Peelle suggested in a comment, most of the stimuli in these tests seem to be weighted toward visually-based concepts (camels, pyramids, etc.), ie, things you can draw or make picture judgments on. If similar semantic deficits could be demonstrated in a modality that is not so visual, auditory, or language based, the case for a truly amodal deficit would be much stronger.
In support of this idea are anecdotal observations that SD patients, despite severe semantic deficits, can nonetheless engage in hobbies, cook, and demonstrate correct object use even for objects that they fail to name or provide correct semantic judgments on.
Hodges et al. (2000, Brain, 123:1913-25) put these anecdotal observations to a rigorous test in an attempt to provide evidence for the amodal semantic deficit idea. They asked SD patients to demonstrate the use of various real objects. Patients were impaired, and performance on the object use task was correlated with the patient's semantic assessment measures. The authors conclude that the patients' deficits arise from disruption of a common semantic system that affects recognition, naming, ... and object use. They explain the apparent preservation of object use in anecdotal reports to "affordances, problem solving, available context, etc." p. 1923.
Does this prove that the deficit is amodal? No.
Problem #1: The task has a strong visual component.
If they are having trouble recognizing the objects because of degraded visual-object representations, they might have trouble knowing what to do with them. But, you might ask, they are touching the objects too, why can't they recognize by touch? They should be able to, but only if the objects were fully explored tactilely (and it is not clear they were), and are familiar to the subject by touch alone. If you look through the list of objects, many of them are very similar tactilely on their grasping region: bottle opener, toothbrush, pencil, whisk, hairbrush, spoon, hammer, potato peeler are all objects that have handles and would be difficult to differentiate if only the handle was grasped. Other objects, such as a pencil sharpener and a corkscrew (depending on the type) -- two objects that generated many errors -- may be hard to identify by touch.
But if this is right, wouldn't patients be expected to have the same trouble in their hobbies and everyday object use? Yes, they should have the same trouble visually, but as Hodges et al. point out, there is context to aid in constraining object use. And this brings us to:
Problem #2: They weren't actually asked to use the objects!
Despite what I said about visual recognition being a possible source of the deficit, I think the fact that patients were not actually asked to use the objects is the main problem with this study. Here is a description of the methods: "Subjects were given each object ... in isolation and asked to demonstrate its use." p. 1916 As there were no bottles to be opened, pencils to be sharpened, ingredients to whisk, potatoes to peel, etc., patients effectively had to quasi-pantomime the actions with object in hand. This differs from actually using the objects in a real context such as cooking, and differs from the investigators' "novel object test" where they had to actually use novel tools in the performance of a task (SD patients were not impaired on this task).
A much better design would have been to give subjects a task to actually perform on each trial (opening a wine bottle, whisking an egg) and giving them three objects to choose from just as in the novel object test. I bet performance would have been much better!