A lamp starts off. After half a minute it is switched on. After a further quarter of a minute, three quarters in all, it is switched off. After a further eighth it is switched on, and so on, each switch coming after half the time that remains. The minute has infinitely many switches in it. Thomson's question is: at the end of the minute is the lamp on or off?
- At each switch the lamp changes state, from off to on or on to off.
- There are infinitely many switches, all over before the minute is up.
- At the end of the minute the lamp is on or off.
- It cannot be on, since every time it was switched on it was later switched off; and it cannot be off, by the same argument with the switches reversed.
- So at the end of the minute the lamp is neither on nor off, against the third premise.
The argument ends where no one can stay: a lamp that is neither on nor off. Thomson drew the conclusion that the setup is impossible, which is to deny the second premise. It is worth seeing what the setup does and does not fix before deciding that.
The lab's lamp setting shows both sequences. The switches happen at the partial sums of the halves, so the nth switch is at time 1 − 1/2^n of the minute. After ten switches the lab reports the time 1023/1024, the same figure as in “Zeno's Dichotomy”. Those times converge to 1.
time of switch n: 1 − 1/2^n
state after switch n: on, off, on, off, …
The second sequence is not a number that approaches anything. The term tile shows the state after n switches: with ten switches it reads off, and with eleven on. Odd gives on and even gives off, because the lamp began off and each switch reverses it. The terms never settle toward either, so there is no limit state in the sense that there was a limit time. The second preset is eleven switches: the state flips while the time moves only from 1023/1024 to 2047/2048.
Now look at premise four. The setup supplies a state for every moment before the end of the minute: at each time 1 − 1/2^n there is a definite state. It supplies none for the end of the minute, because that moment is not the time of any switch, and the rule says only what the lamp does at each switch. A rule about how the state changes at each step, applied to every step, says nothing about what follows all of them. “Every time it was switched on it was later switched off” is true of every moment before the minute and silent about the minute itself, and the fourth premise reads the rule as if it fixed a state there.
That is one way to reject Thomson's conclusion, and it is the standard one: exit one, against the fourth premise. The story is consistent, because its description of the lamp before the minute is complete and consistent and does not mention the state at the minute. Whatever the lamp is doing then, it does not violate the description. A rule that did say, such as “the lamp is on at the end of the minute”, would be a further stipulation by the one who tells the story, and either choice is consistent with the rest. The price is that a story can describe a lamp at every moment of a minute but one and owe nothing about that one.
The case against this reply has force too. A physical lamp has a state at every moment, and if the setup describes a possible lamp then the end of the minute has a state that the setup does not say. A reader who finds that unacceptable has concluded, with Thomson, that the story describes nothing possible. That is exit one as well, against the second premise, and its price is saying which law forbids a body to be switched in this way. The two replies deny different premises and pay different prices, and the lesson's claim is the narrower one the lab can check: the series fixes the time and not the state.