An $\varepsilon$-TRANSITION in an NFA allows the automaton to

1. NFAs can have $\varepsilon$-transitions: $\delta(q, \varepsilon)$ — moves from state $q$ to another state(s) WITHOUT reading any input. 2. These allow the automaton to take 'free' moves between states, useful for building larger NFAs from smaller ones (e.g. for union, concatenation, star). 3. The $\varepsilon$-closure of a state $q$ is the set of all states reachable from $q$ using zero or more $\varepsilon$-transitions. 4. NFAs with $\varepsilon$ are equivalent in power to those without (and to DFAs). 5. Options A, C, D mischaracterise $\varepsilon$-moves. _Source: Jeff Erickson, "Models of Computation", §2.7 ($\varepsilon$-transitions)._