332
P.E. Nachtigall et al.
sured. Another series begins with a magnitude that is lower than the subject
can detect and increases until the subject begins to detect the signal. These
ascending and descending series are presented several times, usually in
random order, and the threshold is defined as the average level at which
the subject switches from reporting present to reporting absent on the
descending series and the level at which the subject switches from reporting absent to reporting present on the ascending series.
In studies of animal psychophysics it is more common to employ a variation on the staircase procedure (Schusternman 1980), which is sometimes
called a modified method of constants or adaptive tracking procedure.
On half of the trials a signal is presented. On the other half of the trials
the signal is omitted. These signal-absent trials are intended to prevent the
animal from being biased to respond present on each trial. A correct
response on each signal-present trial reduces the magnitude of the signal
by a fixed amount and each incorrect response on a signal-present trial
causes the magnitude of the signal to increase. Each time the change in the
signal magnitude reverses sign (i.e., from increasing to decreasing or from
decreasing to increasing) is called a reversal. The threshold is defined as the
average signal level at which these reversals occur. This method is different
from the standard adaptive staircase procedure, in that it includes use of
signal-absent or catch trials, as is used in the method of constants (see
below).
A number of variants of this procedure have been used. Sometimes two
positive responses in a row are required before the signal is decreased, but
only one signal-absent response is required to raise the magnitude. This
variant is called the one-up/two-down procedure. Sometimes, the first few
trials of a day start with high signal magnitudes, which are then decreased
in large steps until the first reversal, and then modified by small steps after
that.
This procedure is very efficient in obtaining thresholds because it allows
the investigator to quickly narrow the range of magnitudes being investigated. Often only a few trials are necessary to find the reversal point. The
procedure sometimes presents pragmatic difficulties, however, because the
signal is always manipulated to be near the subject's threshold, and the difficulty of the task can be frustrating to the subject.
The method of constants is another standard procedure that has been
widely used in animal psychophysical studies. The experimenter chooses
several (often five to nine) different signal levels that are expected to
bracket the subject's threshold (i.e., some are expected to be above and
some below). The lowest signal levels are chosen to be almost never
detected and the highest levels are chosen to be almost always detected.
The actual magnitudes that are used are often selected using a simplified
staircase procedure, but they may be estimated on the basis of some theory
or past experience with this signal. A group of trials is presented containing both signal-present and signal-absent trials. All of the signal-present
P.E. Nachtigall et al.
sured. Another series begins with a magnitude that is lower than the subject
can detect and increases until the subject begins to detect the signal. These
ascending and descending series are presented several times, usually in
random order, and the threshold is defined as the average level at which
the subject switches from reporting present to reporting absent on the
descending series and the level at which the subject switches from reporting absent to reporting present on the ascending series.
In studies of animal psychophysics it is more common to employ a variation on the staircase procedure (Schusternman 1980), which is sometimes
called a modified method of constants or adaptive tracking procedure.
On half of the trials a signal is presented. On the other half of the trials
the signal is omitted. These signal-absent trials are intended to prevent the
animal from being biased to respond present on each trial. A correct
response on each signal-present trial reduces the magnitude of the signal
by a fixed amount and each incorrect response on a signal-present trial
causes the magnitude of the signal to increase. Each time the change in the
signal magnitude reverses sign (i.e., from increasing to decreasing or from
decreasing to increasing) is called a reversal. The threshold is defined as the
average signal level at which these reversals occur. This method is different
from the standard adaptive staircase procedure, in that it includes use of
signal-absent or catch trials, as is used in the method of constants (see
below).
A number of variants of this procedure have been used. Sometimes two
positive responses in a row are required before the signal is decreased, but
only one signal-absent response is required to raise the magnitude. This
variant is called the one-up/two-down procedure. Sometimes, the first few
trials of a day start with high signal magnitudes, which are then decreased
in large steps until the first reversal, and then modified by small steps after
that.
This procedure is very efficient in obtaining thresholds because it allows
the investigator to quickly narrow the range of magnitudes being investigated. Often only a few trials are necessary to find the reversal point. The
procedure sometimes presents pragmatic difficulties, however, because the
signal is always manipulated to be near the subject's threshold, and the difficulty of the task can be frustrating to the subject.
The method of constants is another standard procedure that has been
widely used in animal psychophysical studies. The experimenter chooses
several (often five to nine) different signal levels that are expected to
bracket the subject's threshold (i.e., some are expected to be above and
some below). The lowest signal levels are chosen to be almost never
detected and the highest levels are chosen to be almost always detected.
The actual magnitudes that are used are often selected using a simplified
staircase procedure, but they may be estimated on the basis of some theory
or past experience with this signal. A group of trials is presented containing both signal-present and signal-absent trials. All of the signal-present