177
responses in a row which makes the stimulus less detectable and the one nonresponse makes the stimulus more detectable. It is important to determine the most
appropriate framework with respect to the number of presented trials (fi xed or variable), threshold characterization (discussed later in this chapter), and psychometric
model used to estimate threshold (Garcı́ a-Pérez 1998 ). The staircase methods can
be a very useful tool in behavioral methods in cases where trial number is restricted
because the adaptive procedure mainly samples at or near threshold. This allows for
the greatest confi dence around the threshold estimate and reduces the number of
trials by down sampling at the tails of the psychometric function. The staircase
method is only effective for binomial response choices and is problematic in use for
multinomial responses (for example, in operant feeding assays with multiple locations). Furthermore, real-time feedback and analysis about behavioral outputs are
required for adaptive tracking. That is, experimenters are required to analyze
whether the response outcome is positive or negative in the time between stimulus
presentations.
4.4.2 Other Adaptive Procedures
Other adaptive stimulus presentation procedures like QUEST (Watson and Pelli
1983 ) and ML-PEST (Harvey 1996 ) can also be used in behavioral testing (as
reviewed by Treutwein 1995 ). Unlike the previously discussed methods, these procedures rely on a Bayesian framework and use an estimation of threshold (from
electrophysiological data, for example) as a prior probability of stimulus detection
and weight stimulus presentations around the previously estimated threshold. These
adaptive procedures could be useful in behavioral studies with fi sh species that have
low evoked response rates to acoustic stimuli.
4.5 Threshold Criteria
Auditory threshold criteria can vary widely across behavioral studies and are an
important consideration when comparing thresholds between different species or
conditions (Hawkins 1981 ). Because there is a continuous distribution of response
probability, accurate characterization of a threshold is important to understanding
the auditory sensitivity and capability of animals. However, because we do not fully
understand the mechanisms underlying how the probability of auditory evoked
responses changes across animals, auditory thresholds determined using different
criteria are very diffi cult to compare within and across species.
Auditory thresholds obtained using the method of constant stimuli often have the
greatest variability and are least accurate in threshold characterization (Fay 1974;
Lu et al. 1996 ; Yan and Popper 1993 ; Zeddies and Fay 2005 ). Thresholds obtained
by this method are often defi ned as a 0.5 probability of response, a positive response
greater than three standard deviations above the mean expected response probability
Revisiting Psychoacoustic Methods for the Assessment of Fish Hearing
responses in a row which makes the stimulus less detectable and the one nonresponse makes the stimulus more detectable. It is important to determine the most
appropriate framework with respect to the number of presented trials (fi xed or variable), threshold characterization (discussed later in this chapter), and psychometric
model used to estimate threshold (Garcı́ a-Pérez 1998 ). The staircase methods can
be a very useful tool in behavioral methods in cases where trial number is restricted
because the adaptive procedure mainly samples at or near threshold. This allows for
the greatest confi dence around the threshold estimate and reduces the number of
trials by down sampling at the tails of the psychometric function. The staircase
method is only effective for binomial response choices and is problematic in use for
multinomial responses (for example, in operant feeding assays with multiple locations). Furthermore, real-time feedback and analysis about behavioral outputs are
required for adaptive tracking. That is, experimenters are required to analyze
whether the response outcome is positive or negative in the time between stimulus
presentations.
4.4.2 Other Adaptive Procedures
Other adaptive stimulus presentation procedures like QUEST (Watson and Pelli
1983 ) and ML-PEST (Harvey 1996 ) can also be used in behavioral testing (as
reviewed by Treutwein 1995 ). Unlike the previously discussed methods, these procedures rely on a Bayesian framework and use an estimation of threshold (from
electrophysiological data, for example) as a prior probability of stimulus detection
and weight stimulus presentations around the previously estimated threshold. These
adaptive procedures could be useful in behavioral studies with fi sh species that have
low evoked response rates to acoustic stimuli.
4.5 Threshold Criteria
Auditory threshold criteria can vary widely across behavioral studies and are an
important consideration when comparing thresholds between different species or
conditions (Hawkins 1981 ). Because there is a continuous distribution of response
probability, accurate characterization of a threshold is important to understanding
the auditory sensitivity and capability of animals. However, because we do not fully
understand the mechanisms underlying how the probability of auditory evoked
responses changes across animals, auditory thresholds determined using different
criteria are very diffi cult to compare within and across species.
Auditory thresholds obtained using the method of constant stimuli often have the
greatest variability and are least accurate in threshold characterization (Fay 1974;
Lu et al. 1996 ; Yan and Popper 1993 ; Zeddies and Fay 2005 ). Thresholds obtained
by this method are often defi ned as a 0.5 probability of response, a positive response
greater than three standard deviations above the mean expected response probability
Revisiting Psychoacoustic Methods for the Assessment of Fish Hearing