and with the computer, Turing asserted that the computer should be judged
‘‘intelligent.’’ It would then be said to have ‘‘passed Turing’s test.’’
Note that Turing’s test is a strictly behavioral test because the interrogator
has no information about the physical attributes of the subject, but only about
its behavior. In the original version, this behavior is strictly verbal, but there is
no reason in principle why it needs to be restricted in this way. The interroga-
tor could as kthe subject to draw pictures or even to carry out tas ks in the real
world, provided the visual feedbac kthe interrogator received did not provide
information about the physical appearance of the subject.
The same imitation game can be used for deciding about the appropriateness
of any other cognitive description, including whether the subject is ‘‘conscious.’’
Again, simply asking the subject whether it is conscious will not discriminate
betweenthemachineandapersonbecausethemachinecaneasilybepro-
grammed to answer that question in the affirmative. Similarly, appropriate re-
sponses to questions asking it to describe the nature of its visual experiences or
pain experiences could certainly be programmed. But even if they could, would
that necessarily mean that the computer wouldbeconscious or only that it
wouldactasifitwereconscious?
If one grants that physical appearance should be irrelevant to whether
something is conscious or not, Turing’s test seems to be a fair and objective
procedure.Butitalsoseemsthatthereisafactatissuehereratherthanjustan
opinion—namely, whether the target object is actuallyconsciousor merely sim-
ulating consciousness—and Turing’s test should stand or fall on whether it
gives the correct answer. The problem is that it is not clear that it will. As critics
readily point out, it cannot distinguish between a conscious entity and one that
only acts as if it were conscious—an automaton or a zombie. To assert that
Turing’s test actually gives the correct answer to the factual question of con-
sciousness, one must assume that it is impossible for something to act as if it is
conscious without actually being so. This is a highly questionable assumption,
although some have defended it (e.g., Dennett, 1991). If it is untrue, then pass-
ing Turing’s test is not a sufficient condition for consciousness, because autom-
atons can pass it without being conscious.
Turing’s test also runs into trouble as a necessary condition for conscious-
ness. The relevant question here is whether something can be conscious and
still fail Turing’s test. Although this might initially seem unlikely, consider a
person who has an unusual medical condition that disables the use of all the
muscles required for overt behavior yet keeps all other bodily functions intact,
including all brain functions. This person would be unable to behave in any
way yet would still be fully conscious when awake. Turing’s test thus runs
afoul as a criterion for consciousness because behavior’s lin kto consciousness
can be broken under unlikely but easily imaginable circumstances.
We appear to be on the horns of a dilemma with respect to the criteria for
consciousness. Phenomenological criteria are valid by definition but do not ap-
pear to be scientific by the usual yardsticks. Behavioral criteria are scientific by
definition but are not necessarily valid. The fact that scientists prefer to rely on
respectable but possibly invalid behavioral methods brings to mind the street-
light parable: A woman comes upon a man searching for something under a
streetlight at night. The man explains that he has lost his keys, and they both
Visual Awareness 17