In our laboratories we have used a positive reinforcement technique
to measure the psychophysical performance of the chinchilla in several auditory
experiments. Using this technique we have shown that chinchillas require larger
signal-to-noise ratios for detection of masked tonal signals than human subjects do
(Niemiec et al., 1992). We have used narrowband noises, notched-noises, and rippled noises
to measure the auditory filter functions; the bandwidth of auditory filters for
chinchillas are similar to those of human subjects (Niemiec et al., 1992). We have shown
that for chinchillas, wideband and narrow band noise increment thresholds vary as a
function of frequency and level in approximately the same way as they do for human
subjects (Shofner et al., 1993; Shofner and Sheft, 1994). Recently, we have shown that
chinchillas discriminate among stimuli that give rise to complex pitch in humans in about
the same way as human subjects discriminate among these same stimuli (Shofner and Yost,
1995; 1997). In this section, we describe the psychophysical procedure used in our animal
behavioral lab.
Behavioral Shaping and Training
A naive animal is weighed daily over a period of several days in order to establish a
baseline body weight. After a baseline weight has been obtained, the animal is placed on a
restricted diet of 10 g daily of Purina Chin Chow. The body weight of the animal is
gradually reduced to be approximately 85% of the baseline weight.
The naive chinchilla is placed inside the cage in the testing booth. The cage measures 16"l x 10"w x 12"h. The animal is not restrained in any manner and is free to roam around in the cage. At one end of the cage is a reward chute attached to a response lever. A Gerbrands pellet dispenser is located outside of the cage and drops Noyes food pellets (Formula N, 20 mg pellet) into the reward chute. A loudspeaker is located approximately 6 inches in the front of the animal at approximately 30o to the right of center.
As the animal roams around exploring the inside of the cage, the experimenter watches through a glass window. At this point, the experimenter has contol of the pellet dispenser. When the animal gets near the response lever/reward chute, the experimenter activates the pellet dispenser to drop a food pellet into the reward chute. Each time the animal comes near the reward chute, the experimenter dispenses a food pellet. After a short time, the chinchilla learns that this is where the food is at and will generally approach the reward chute and wait for a pellet. At this point, the experimenter does not activate the pellet dispenser. When the chinchilla fails to receive a food pellet, it will generally place its face in the reward chute, and by doing so, it presses the response lever down. When the response lever is pressed down, a micro switch is closed causing a food pellet to be dispensed and a test signal to be played over the loudspeaker. Thus, the animal learns that when the response lever is pressed down, a food pellet is dispensed. It should be noted that when the animal eats the pellet, it releases the response lever. This phase of training is generally carried out over a period of days until the animal is consistently pressing the response lever down.
The next phase of training is where the animal learns to associate
receiving a food reward with the presence of the auditory test signal. In this phase, when
the animal presses down on the response lever, the lever must be held down for a duration
of 1 s. After the 1 s holdtime, the test signal is played through the loudspeaker, and if
the animal releases the lever, a pellet is dispensed. While this procedure appears to be
very different from the initial behavioral shaping procedure described above, it is
virtually not that different for the animal, and the chinchillas quickly learn how to do
this. After the animal learns to do this procedure consistently, the holdtime is increased
to 2 s. This is often more difficult for the chinchilla to learn. Over the course of
several months, the holdtime is gradually increase to 8 s. After the animal is working
consistently with the fixed holdtime, the holdtime is then varied randomly for each trial.
Finally, blank (i.e. nonsignal) trials are gradually added. For a blank trial, the animal
should continue to hold the response lever down for a correct response. If the animal
continues to hold the response lever down during a blank trial, then a food pellet is
dispensed as a reward.
Psychophysical Procedure
we will illustrate the psychophysical procedure by using the study of Shofner and Yost
(1995) in which chinchillas discriminated an infinitely-iterated rippled noise (IRN) from
a flat-spectrum wideband noise (WBN). Figure 1 illustrates the paradigm. The paradigm is a
modified "yes/no" procedure in which the animal discriminates a test sound from
a standard sound.
Wideband noise bursts (500 ms with 10 ms rise/fall times) were presented continually once per second throughout the testing session regardless of whether the animal initiates a trial. The wideband noise is the standard sound in this example. The animal initiates a trial by pressing down on a response lever. After a trial was initiated, the noise bursts continued for 1-5 bursts for each trial. These additional 1-5 noise bursts result in a holdtime of 1150-5150 ms. The number of noise bursts that continue after a trial is initiated varies randomly for each trial and is determined from a rectangular probability distribution. If the animal releases the lever before the random holdtime, the procedure is halted and the computer waits for the animal to reinitiate the trial by pressing the lever. If the animnal holds the response lever down for the duration of the holdtime, then either a signal trial (IRN) or a blank trial (i.e. WBN) was presented. A signal trial consists of two bursts of IRN, which is the test sound; a blank trial consists of two additional bursts of the WBN, which is the standard sound. The response window is coincident with the duration of the signal/blank trial, but begins 150 ms after the onset of the first IRN/WBN burst and lasts until the onset of the next WBN burst; consequently, the duration of the response window is 1850 ms.
Whenever the animal releases the lever during the response window, the release is scored as a "yes" response. That is, a lever release is the animal's way of saying "Yes, I hear the test sound." If the animal continues to hold the lever down throughout the response window, the response is scored as a "no"; that is, "No, I do not hear the test sound." If the animal releases the lever during a signal trial, then this "yes" response is treated as a hit, while a lever release during a blank trial is treated as a false alarm. If the animal continues to hold the lever down for the duration of the response window, the response is treated as a correct rejection. Chinchillas are rewarded with food pellets for both hits and correct rejections. Time outs are not given for incorrect responses
We use the Method of Constant Stimuli to generate psychometric functions (see Figure 1). Animals run in a block of 40 trials. A block contains 9 different signal levels. For this rippled noise experiment the signal levels are in terms of gain settings. The gain determines the spectral depth of the ripple and the strength of the pitch of the rippled noise. A block of 40 trials consists of 4 trials each at gains of -1, -2,- 3, -4, -5, -6, -7, -8 dB, and 8 trials at a gain setting of -100 dB. (The gain settings produce IRNs having different spectral modulation depths such that for -1 dB there is the largest depth of modulation, and for -100 dB the flat-spectrum WBN is generated). The values of gain are presented randomly for each block. The final 9 point psychometric functions are generally based on a minimum of 50 blocks, or 2000 total trials. These psychometric functions can be fit with logistic or Weibull functions if desired. In our studies, we generally express behavioral performance in terms of d' rather than in terms of p("yes"). Using this Method of Constant stimuli, we typically reward chinchillas 1 food pellet for hits and 2 pellets for correct rejections. The advantage of this procedure is that the entire psychometric function can be generated. However, a disadvantage of this procedure is that it is rather inefficient and time consuming. We have also used a modification of this procedure in an adaptive tracking paradigm, which is much more efficient in terms of data collection (Niemiec et al., 1992; Shofner et al., 1993; Shofner and Yost, 1994).
 |
Figure 1. Behavioral paradigm used for the chinchilla
psychophysical studies. This example illustrates the discrimination of an iterated rippled
noise (IRN) from a flat-spectrum wideband |
In order to assess the validity of this procedure, we have obtained psychometric functions
from human listeners using the animal behavioral procedure. We have measured psychometric
functions for human subjects using this animal behavioral paradigm for discriminating 1
iteration of rippled noise from wideband noise. This discrimination has been carried out
previously by several researchers using force-choice procedures. We define threshold from
the psychometric function as the gain that produces a d' of 1. Figure 2 shows that
thresholds obtained from human subjects in the animal behavioral paradigm (red symbols)
are similar to those obtained with froced-choice procedures (blue symbols). Thus, this
behavioral paradigm appears to provide valid estimates of sensitivity.
 |
Figure 2. Thresholds for the discrimination of rippled noise from wideband noise obtained from human listeners. Blue symbols show thresholds obtained from previous studies using forced-choice procedure. Red symbols show thresholds obtained human subjects using the animal behavioral paradigm illustrated in Figure 1. |
Figure 3 shows d' as a function of gain for 4 human subjects (blue symbols) and 6 chinchillas (red symbols) for the discrimination of infinitely-iterated rippled noise from flat-spectrum wideband noise. The psychophysical data for the chinchillas is shifted to higher gains than that of the human data, suggesting that chinchillas are not as sensitive as human subjects in the discrimination. One question often raised is whether this reflects the true sensitivity of the chinchillas or whether it is merely due to the higher false alarm rates of chinchllas compared to those of human listeners. If the human performance is computed using the largest value of the chinchilla false alarm rates, then human listeners still show more sensitivity than chinchillas. Thus, data derived from this technique appear to produce reliable and valid estimates of the chinchilla's ability to detect auditory signals and to discriminate changes in these signals.
 |
Figure 3. Performance of human listeners and chinchillas for the discrimination of infinitely-iterated rippled noise from wideband noise. Data from human subjects was obtained using the animal behavioral paradigm. Performance is expressed in terms of d'. Red symbols show d's for chinchillas; blue symbols show d's for human subjects; green symbols show the predicted d's for human subjects based on the largest false alarm rate obtained from the chinchillas. |
To hear a 1 s sample of flat-spectrum wideband noise, click here 
To hear a 1 s sample of infinitely-iterated rippled noise at a delay of 4 ms (250 Hz
pitch) with a
gain of -1 dB, click here
The discrimination threshold for the average chinchilla is at a gain -6 dB for
infinitely-iterated rippled noise with a 4 ms delay. To hear a comparison of wideband
noise and this infinitely-iterated rippled noise, click here
The discrimination threshold for the average human subject in the animal behavioral
paradigm is at a gain of -21 dB for infinitely-iterated rippled noise with a 4 ms delay.
To hear a comparison of wideband noise and an infinitely-itereated rippled noise with a
gain slightly above the discrimination threshold (-18 dB), click here The pitch strength of this rippled noise is very weak;
consequently this rippled noise has a very weak hollow sound. The subtle difference in
these two sounds may be best heard over headphones.
Bibliography
Niemiec, A.J., Yost, W.A., & Shofner, W.P. (1992) Behavioral
measures of frequency selectivity in the chinchilla. Journal of the Acoustical Society of
America, 92, 2636-2649.
Shofner, W.P., Yost, W.A., & Sheft, S. (1993) Increment detection of bandlimited
noises in the
chinchilla. Hearing Research, 66, 67-80.
Shofner, W.P. & Sheft, S. (1994) Detection of bandlimited noise masked by wideband
noise in the
chinchilla. Hearing Research, 77, 231-235.
Shofner, W.P. & Yost, W.A. (1995) Discrimination of rippled-spectrum noise from
flat-spectrum
wideband noise by chinchillas. Auditory Neuroscience, 1, 127-138.
Shofner, W.P. & Yost, W.A. (1997) Discrimination of rippled-spectrum noise from
flat-spectrum
noise by chinchillas: Evidence for a spectral dominance region. Hearing Research
110: 15-24.