The other day, I was looking for a description of how Hebrew verbs conjugate, and found an article entitled Tense and Agreement in Agrammatic Production: Pruning the Syntactic Tree by Na'ama Friedmann and Yosef Grodzinsky (185k PDF).
I had a look at it anyway, and found it rather interesting. They reported the case of an older lady who had been suffering from Broca's aphasia for several years. This impaired her ability to produce certain kinds of sentences (including WH-questions, and embedded clauses) and to produce some verb inflections.
Interestingly enough—and probably a big reason why the paper was published—, she was able to inflect verbs for agreement (person, number, and gender) but not for tense, whereas from reading the paper, my impression was that previous work on the topic of agrammatic people implied that all inflection was typically impaired.
So for her, she might say "Etmol ha-yeled kotev" or "Etmol ha-yeled yiktov" ("Yesterday, the boy writes" or "Yesterday, the boy will write") with incorrect tense, but she would never say "Etmol ha-yeled katvu/katva/katavti" ("Yesterday, the boy wrote-PL/wrote-F/wrote-1") with incorrect number, gender, or person. (Or in "other words", she might say "Il-bieraħ, it-tifel jikteb" but not "Il-bieraħ, it-tifel kitbu/kitbet/ktibt".)
The paper describes a number of tests they ran on her, both on comprehension and production, and surmised how the brain may produce syntactically-correct sentences, and how impairment at a certain point will result in problems that involve that point and points further up, and that people with varying severity of the syndrome will vary primarily in the point at which problems occur (meaning, among other things, that there are only a few basic kinds of the syndrome since there aren't that many points where impairment can happen).
Some of it was over my head since I'd never studied linguistics formally, but I still found it an interesting read.
Incidentally, her "input route" (as they called it) appeared to be relatively intact, so she could recognise incorrect sentences when she saw them (including incorrect sentences she had produced herself). The authors also mentioned that in one exercise (where she had to put the negative marker either before or after the copula, as required by Hebrew syntax depending on the tense), she performed badly, but after being taught to try all possibilities and look at the resulting sentence to see whether it was correct, her success rate went up to 100%—because then her input was involved in checking the result, which worked, whereas the built-in checking in her output route was impaired.
What I also found surprising was that apparently, she also had difficulty in reading out sentences she could not have produced (e.g. embedded sentences).