Andy's answer is spot-on if this circuit operates in isolation.
But if there happens to be a positive bias current source connected to Vo+, and a negative current source connected to Vo-, then it is something different. (For simplicity, these bias current sources can be resistors connected to +V and -V respectively.)
Then the outputs form three parallel lines, about 0.6V apart, with the central one (Vo = -Vi).
Vo+ and Vo- may be used to drive the bases of two complementary emitter followers, providing a crude power amplifier. Their emitters are connected together, and each base-emitter voltage is cancelled by the voltage across each diode. Therefore the final output (at the emitters) is approximately Vo = -Vi. But only approximately : errors here lead to crossover distortion.
simulate this circuit – Schematic created using CircuitLab
In its current form, it's not very efficient : if you think about what happens when VO+ = V+/2 you'll see that it runs out of bias current and starts clipping (with the current resistor values), but this illustrates the basic principle, and you'll often see something like this in an audio power amplifier.
Further circuitry is usually added to fix the defects in this basic configuration.