This came from a conversation with leckysupport, and I thought I'd post it here.
Te and Fe monitor changes in individual objects. Te monitors external, directly observable changes in individual objects, while Fe monitors internal, inferred changes in individual objects. They do this one at a time, without reference or comparison to other objects.
Si and Ni monitor changes in relationships between objects. Si monitors changes in explicit, directly observable relationships between objects, while Ni monitors changes in inferred relationships between objects which are not directly observable. Long periods of time are not directly observable, and thus fall under Ni. Short periods of time are directly observable, and so fall under Si.
So, when Te sees a bowling ball rolling along, it sees that the ball is rotating at X revolutions per minute, and travelling at Y speed, etc. Si sees how the bowling ball interacts in directly observable ways with other objects, such as the floor and pins. It sees how the bowling ball bounces off the floor, creating a Te effect in the floor (the floor flexes slightly). It sees how the ball impacts the pins, creating a Te effect in the pins (the pins go flying).
Ni, on the other hand, sees how the bowling ball interacts in ways that are not directly observable with other objects. It sees how the bowling ball interacts with the floor over the long term, creating a Te effect (the ball wears down the finish on the floor). It sees how the bowling ball interacts with the pins symbolically (which is of course not directly observable), drawing comparisons to a cannonball knocking down soldiers, or mankind cutting down forests, etc.
I'm wandering from my point somewhat. To sum up, Te and Fe monitor changes in individual objects, and Ni and Si monitor how those changes affect other objects. Si monitors how objects interact in observable ways (limiting it to short-term interactions), while Ni monitors how objects interact in non-observable ways (including but not limited to interactions between objects separated by long stretches of time or space, or symbolic interactions, etc.).
The following section I'm not certain about, and I would appreciate input from Ni and Si Egos:
Hypothetically, the "future version" of a present object could be considered as a separate object. The interactions between a present object and a "future object" which are separated by too much time for the interactions to be immediately apparent, would be considered with Ni. The interactions between a present object and a "future object" which is close enough in time for the interactions to be immediately apparent, would be considered with Si. In both cases, the changes in the present and future objects themselves are considered with Te (or Fe, as the case may be).
The same would be true of "past versions" of objects, as well as objects separated in space by large or small distances.