Depends on what you mean by pumping air in and out because we have two situations that describe that.
-Just killing fuel to two injectors.
-Killing fuel to two injectors and pulling the spark plug/injector.

According to Honda, aggressive cylinder deactivation results in a 30% increase in fuel economy, which corresponds to a 65% reduction in pumping losses. They do this by deactivating two cylinders, which results in the other two pulling in twice as much air, to make the same power, and a halving of pumping losses, as well as the closure of the two deactivated cylinders, which theoretically halves pumping losses again. Ideally, we'll see a 75% reduction in pumping losses, a half halved. Real world, those two cylinders have to pull in more air, and the two closed ones aren't perfectly closed, so we see a 65% reduction in pumping losses.

So, if we just cut fuel to two cylinders, we theoretically halve pumping losses, and realistically cut them by ~40%. We then have two choices...

-Close the cylinders completely.
-Open the cylinders as much as possible.

If we close them we know we can see a ~65% reduction in pumping losses, but, most cars would need a very high idle to run like this, or different EFI routines (turning the cylinders off past idle) to have the car run in this state. You said the Tempo would barely start, then die. It may start and run if you raise the idle enough, but then again, what's the point of halving pumping losses just to idle at twice the normal speed. That's going nowhere fast. So imo, closing the cylinders is best left up to a manufacturer.

We can also permanently kill the fuel to the two cylinders, and pull out the spark plugs/injectors. This will reduce pumping losses from those two cylinders, but not eliminate them like the air spring system would. They'll still pull air in, and push it out. But, if the two active cylinders still aren't filling completely when cruising at around 40mph in top gear, the two that allow free passage of air will be.

The reason why gasoline engines are inefficient at low loads is the difference between cylinder pressure and crankcase pressure. Diesels don't have this since the cylinder is always completely filled, so they get much better mileage at lower speeds. Opening the spark plugs/injectors allows the pressure to equalize in the two cylinders that aren't active. While pulling air in, and pushing it out, does result in pumping losses, these are very small, and akin to what diesels see. So, I'm guessing that w/o plugs/injectors, pumping losses may be reduced by ~60%, compared to 65% with a totally closed cylinder.

The thing is, if we're going this far, we might as well just pull the pistons to get rid of friction too (which should result in a greater than 30% efficiency increase). So for cylinder deactivation (a bit less than 20%), I think an all or fuel cut are the best approaches. The thing I want to stress is that when people talk about pumping losses in gasoline engines, they usually mean the difference in cylinder/crank pressure, not the movement of air in/out. While both result in losses, the pressure difference is why gasoline engines are so lousy at low load. It's also why diesels are good at low load, since they always pull everything into the cylinder they can.

Temponut, did you get black smoke when you pulled two injectors?

you could easly fix the problem with the o2 sensor by getting an exaust header that has an o2 sensor in one of the exaust streams that is getting fuel, the computer assume that all 4 cylenders are getting the same amount of fuel because it's opening each injector the same amount, so it can get by with monitering just one cylenders exaust.
https://www.jaydm.com/assets/dc/4-2-1ss2p.gif

if you want to make this a perminet modification, why not just swap in a smaller engine? it would run alot smoother, but I see having it set up kind of like the vtec-e system, where it operates as a more efficent engine 95% of the time, but when you ask it for more power, how the engine operates shifts, and the power is there.

Well yeah, a turbocharged 700-1000cc engine would probably work great in my car, but it takes a lot more time, money, effort, knowledge, etc. to put in, and it'd ruin any resale value of the car (although if my last car is any indication, I'll be the last owner of this one.) The conversion cost isn't justified unless you drive a lot, and I don't -- my current commute is only 16 miles.

That is a good thought about O2 sensor placement. It may already be on one cylinder, in which case I may be able to just pull the injectors and plugs on the opposite two. I'll need to take a look.

Will removing the plugs from two wires damage the electronic ignition, or would it be better to plug them back in and cap them?

- Bruce

There's a lower limit for doable (no custom motor mounts etc) small engine swaps and I'm trying to figure out whether a gasoline engine can be modded relatively easily to get the same mileage as a diesel during the same driving conditions. Would it better for the average driver to just swap in a diesel, or can they see the same mileage for way less by modding the gasoline engine they have for lean burn/cylinder deactivation?

I believe I've already spotted a car in the Garage where the owner swapped out a gas for diesel and had a huge increase in FE, but I can't remember whose. I know I've seen at least a couple with swapped engines.

A `92 isn't new enough to use a Scangauge...did you verify the decreased FE at the pump?

I'd expect more throttle would be needed, but that doesn't necessarily indicate lower efficiency...a good example is lugging an engine at low speed (great efficiency, and it takes a lot of throttle).

It could be because of the ECU going into closed loop... The difference in mileage for something like a normal versus lean burn civic is ~20% at cruise iirc. So if going from ~15:1 to 17-19:1 allows for this, a richer default map at 12-13:1 may wipe out any gains made by only running on two cylinders. In order to find if I'm completely full of it we'll need an engine running on half the cylinders via a fuel cut with an oxygen sensor in one of the exhaust runners of a running cylinder and a SG to get accurate back to back runs and determine if the ecu is in closed or open loop mode.

I didn't know about the previous post on the subject, but found it today:

https://www.gassavers.org/showthread.php?t=48&page=3

Bottom line was 24.3 MPG with fuel cut to two cylinders vs 26-28 normally.

- Bruce

Running it in a higher gear just to avoid vibrations defeats the purpose of the experiment. That combined with richer running, and it's no wonder the mileage was worse. At the very least, we need a couple runs at the same rpm, ideally with the O2 sensor in an active exhaust runner.