As I understand it, the practice of distinguishing between agent (ergative) and non-agent (absolutive), rather than between nonpatient (nominative) and patient (accusative) as is common in Indo-European languages.

For a 'very' brief overview of different case-marking systems, see http://www.xibalba.demon.co.uk/jbr/ranto/r.html — part of a rant against Esperanto, but you can ignore that here; this appendix focuses only on cases. Ergative languages use approach (E) in his list.

How do we express this in Lojban? gausmi?

• gernrgativiti --And (mi stidi)

Unfortunately English not always marks this distinction (e.g. "return",
the infamous keyword of xruti)... See Ergativity.

Non-Agentive form & it's corresponding Agentive:

(There are doubtless others.)

• The place structure of xruti has changed during its history...
  • A note here says (1994) that xruti was to become non-agentive, although it remains agentive on the gismu lists of 9-6-94.
  • I will have to check my notes, but IIRC the non-agentive meaning was changed to the present agentive at about the same time as the Great Rafsi Shift.

• Lojban distinguishes forms of causation...One could extend the foregoing series with each of the others.

• One must also consider the "hidden doers", those with non-x1 agent places:
  • smuni has meaning-------tersmu interprets
  • tatpi is tired------seltatpi tires
  • fengu is angry------selfengu angers
  • torni is twisting-------seltorni twists
  • xalni is panicked-------selxalni terrifies
  • zgike is music------selzgi makes music

That's a very useful list. We shouldn't overlook the usefulness of zil-, though:

zilsti comes to a stop

sisti halts (or dicra interrupts)

zilxru reverts

xruti restores

zilcatra dies

catra kills

zilcfipu is confused

cfipu confuses

--And

• Wouldn't "is killed" be better as a translation than "dies"? That is, there has to be a killer or there isn't really any catraing involved. I'd use mrobi'o (become dead) for just plain dying "without a cause". (And I'm not sure whether consensus was reached on the usage of zil- in this manner. xu la .andban. lojbo?) --mi'e filip
• It's a reasonable suspicion to have when I say something, but in this instance it's pretty uncontroversial, at least among the jboskepre. (I can't speak for nonjboskpre, since I don't get to hear their relevant views.)