Exposure value, EV, is log2(Fnumber^2/Time_in_secs). That is, if the aperture is set to f/2, the Fnumber is 2; and suppose time is set to 8 secs, log2(2^2/8) = -1 EV. Closing the aperture by 1 stop, that is increasing the Fnumber by sqrt(2), we are getting Fnumber = 2.8. Decreasing the shutter speed to compensate, also by 1 stop, we are getting 16 secs, and the formula above results in the same EV number, -1 EV.

For your shots the EV numbers are:
14.83	15.26	15.31	15.26
That is, the difference between the light in the first shot and in the three others is close to 1/2 stop. Given the light in the scene did not change...

The exposure experiments are better done when the light does not change, say, indoors, with incandescent lights (not fluorescent, due to flicker).

Keep the area under the focus point constant if experimenting with matrix metering, as the metering system gives the priority to the brightness under the focus point. Keep the framing the same. Watch that Active D-Lighting (ADL) is Off, it may cause the results to be all over the place. Any "Auto" may also skew the results.

Drop the exposure triangle, exposure does not depend on ISO. Accepting exposure triangles folks start to think that the reason for the noise is high ISO, not the insufficient amount of light.

Also, exposure has little to do with image brightness, change the monitor brightness and what now? - does exposure change? No. Exposure depends on shutter speed, lens aperture (T-stop, actually) and scene luminance at the time the image was taken.