Once the requisite number of Lojbanists did an undergraduate course in syntax (you may commence throwing darts at effigies of Nick Nicholas at your leisure), it was realized that there were a lot of verbs (brivla) whose place structures contained both a raised concrete noun (usually x1), and an abstraction noun which itself contained the first noun. For example, the place structure of fenki used to be
- x1 is crazy in behavior x2 (abstraction) by standard x3
But any abstraction that would go into x2 would contain the x1 noun: any crazy behaviour would automatically be the behavior of the crazy person. For example, you'd get
- la .djang. cu fenki lonu la .djang. cu dasni lo zirpu
- la .djang. cu fenki lonu la .djang. cu dansu la jipci
- la .djang. cu fenki lonu la .djang. cu tavla bau la .lojban.
The question then became: does the x1 tell us anything the x2 wasn't already telling us? We know who was involved in the crazy behavior, because that person would be a noun inside x2. (More specifically, he or she would be the active party: someone hitting random strangers is crazy; someone being hit by random strangers isn't — although arguably someone allowing themselves to keep being hit by random strangers is.) Was there any reason, then, to grant the person an extra place in the overall bridi? The decision was, no: behavior is what is crazy, so you can work out that the person acting out the behavior is the crazy person. There's no need to have an extra place for the person, when you can already work out who they are. The same conclusion was arrived at for cinri: it is abstractions — events and qualities — that attract interest; and an interesting person is simply a person involved in an interesting abstraction.
All well and good; but natural languages do raising for a reason. So when Lojban has its gismu without raising, it gains in eliminating redundancy and logical muddledness; but it loses in ‘naturalness’. We like talking about people rather than abstractions in our languages; and Lojban should not go out of its way to form an exception to this.
There is a solution of sorts to this problem using tu'a; but it doesn't actually do what raising does in natural languages: it doesn't change the x1 place from an abstraction to a concrete noun. And there are times you will want to do just that.