- Email: [email protected]
The design and calibration of a startle measurement system
Physiology & Behavior, Vol. 36, pp. 377-383. Copyright © Pergamon Press Ltd., 1986. Printed in the U.S.A.
0031-9384/86 $3.00 + .00
BRIEF COMMUNICATION
The Design and Calibration of a Startle Measurement System J A M E S V. C A S S E L L A
AND MICHAEL
DAVIS
Yale University School o f Medicine, and the Ribicoff Research Facilities of the Connecticut Mental Health Center, 3 4 Park St., New Haven, CT 06508 R e c e i v e d 18 J u l y 1984 CASSELLA, J. V. AND M. DAVIS. The design and calibration of a startle measurement system. PHYSIOL BEHAV 36(2) 377-383, 1986.--The present study sought to determine appropriate instrumentation for amplification and calibration of cages used to measure the acoustic startle response in rats. Fourier analysis indicated that the characteristic frequency of the rat startle response is about 5-15 Hz. This value was consistent in cages differing widely in resonant frequency, among several different transducers and across a range of large and small startle responses. Given this characteristic frequency range of startle, it is suggested that amplifiers fitted with band pass filters centered at about 10 Hz should be ideal for measuring startle while simultaneously excluding non-startle activity. A device is described which vibrates startle cages at 10 Hz, since this seems most appropriate for calibrating the sensitivity of a startle system. Data are presented showing that this type of calibrator is more valid than an impact-type calibrator. Startle response
Amplifiers
Calibrators
T H E startle response in rodents is being used increasingly for the analysis of animal behavior. This reflex, which is typically elicited by a loud sound or air puff, has a short, reproducible latency and occurs in virtually every rat. Startle is a highly graded reflex that can be altered by changes in the parameters of the eliciting stimulus, the surrounding environmental stimuli, or general state of the animal [6,11]. During repetitive stimulation, startle habituates or sensitizes, depending on the exact parameters used, and has provided an excellent model system for the analysis of behavioral plasticity. Startle amplitude can be increased by prior fear conditioning [1,7] and currently is being used as a model system to evaluate the pharmacology [4,5] and anatomy [13] of classical conditioning. In addition, startle is sensitive to various drugs and is currently being used to investigate the pharmacology of behavior [6]. To date, many different systems have been described for measuring startle. In most cases, the startle response of the animal results in movement of a cage which is translated into a voltage, amplified, and then displayed, typically as a pen deflection or digital readout. A variety of transducers such as accelerometers [14], magnets within coils [10], photograph cartridges [2,3], strain gauges [15], piezo electric f'dm [12] or even purely mechanical systems [1] have been used. Recently, commercially designed startle cages have become available. A major, unsolved problem in the field of startle measurement is that no one has developed a satisfactory way to calibrate startle cages so that one cage is just as sensitive as another in measuring startle. This is both a problem within laboratories (since it would be desirable to have all cages in a
given laboratory set at a comparable sensitivity) as well as across laboratories (since it may be desirable to know how similar cage sensitivities are in different laboratories). Most calibration systems involve striking a startle cage with a reproducible force and measuring the resultant movement of the cage. F o r example, a solenoid may be positioned in a cage in such a way that when it is activated, the pin hits the cage which generates a voltage at impact. Dropping the pin from various heights or dropping pins of different weights can be used to determine the output linearity of the cage. In other cases, a solenoid is connected to a rigid superstructure. Activation of the solenoid moves an extension rod which hits the cage in a reproducible way [ 10]. Other workers have dropped clay or acrylic balls of different weights from various heights and recorded cage output [7,12]. While it is possible to make calibrators that automatically and reproducibly impact onto startle cages, it has never been clear if impact-type calibrators are valid devices for mimicking the behavior of the animal. In this paper, we will present evidence suggesting that impact-type calibrators may have little validity in this regard. Moreover, we will describe what is believed to be the essential aspect of the startle response dictating both how startle should be measured and how calibrators should be constructed.
The Problem of Amplification and Calibration of Startle Cages F o r the past several years we have been using the startle cage system shown in Fig. 1. Each of 5 stabilimeters consists
377
378
CASSELLA AND DAVIS
FIG. 1. Photograph of a startle cage. Accelerometer looks like a steel nut with a wire coming out of it and is sandwiched between bottom of the cage and a rubber stopper.
of an 8x 15× 15 cm Plexiglas and wire mesh cage suspended within a 25x20x20 cm steel frame. Within this frame the cage is rigidly sandwiched between 4 compression springs above, and a 5 x 5 cm rubber cylinder below. An accelerometer (M. B. Electronics Type 302), with a sensitivity of 40--60 mV/g, is located between the bottom of the cage and the top of the rubber cylinder. Its output is fed to an MB-N504 accelerometer amplifier. Cage movement results in a slight and imperceptible displacement of the accelerometer which generates a voltage that is proportional to the rate of cage displacement. Startle is defined as the peak accelerometer output that occurs within 200 msec after presentation of the startle-eliciting stimulus. Because the cage is rigidly held in place and no mechanical adjustments are required, the system has remained very stable over several years.
These 5 cages were calibrated by dropping a wooden arm attached to a rigid fulcrum onto the cage from a fixed height. By attaching different weights to the arm, graded outputs could be measured and the sensitivities of each accelerometer amplifier adjusted to give comparable outputs for all the cages. Increased research demands required construction of an additional 5 cages. A major change in this new system was to replace the M. B. Type 302 accelerometer, which was no longer available, with an Endevco Type 2217E accelerometer. The amplifier chosen for this new system was a WPIDAM 5 amplifier (WPI Instruments, Inc., New Haven, CT) since the one made by Endevco was too expensive. However, it was discovered that the WPI amplifier differed from the MB-N504 amplifier in that the WPI detected and
STARTLE M E A S U R E M E N T SYSTEM
379 I
A soo ACCELEROMETER
OUTPUT (z W 0
", -
I."
I
m'
t
STIMULUS
l,ovo,,
I0
20 30
40 50 60 70 80 90 100
2000
B 1500
20 MSEC.
BOO
MARKER
10 20 SO 40
50 eo 7o 80 oo 100
to
2o 3o 40 60 SO #o 8o go-~oo
FREQUENCY
FIG. 2. Oscilloscopic tracing of the output of the startle cage using a wide band amplifier (low frequency f'dter= 1 Hz, high frequency=4kHz). The rat was startled by a 20 msec, 4-kHz tone. The stimulus artifact in the beginningof the trace is present because the cage acts like a microphone and picks up the vibration produced by the tone, which is only partially eliminated by the high frequency f'dter. Within about 12 msec of stimulus onset, the cage begins to move down at a frequency of about 10 Hz. The higher, smaller amplitude frequency reflects the resonant frequency of the cage (about 55-60 Hz).
amplified the artifact produced by the 4-kHz tone. Consequently, a low pass f'flter was built into the WPI amplifier so that signals above about 200-Hz were f'dtered out, thus eliminating the stimulus artifact when the 4-kHz tone came on.
Having made these modifications to the WPI amplifier we attempted to calibrate the 5 cages using the MB-N504 amplifier and the 5 cages using the WPI amplifier so that all cages were equally sensitive to vertical displacement. The calibrator consisted of a 28 volt solenoid held rigidly in an aluminum frame that fit snugly into the startle cage. When 28 V was applied for 50 msec across the solenoid core, a 50 g pin was lifted to a height of 15 mm and then dropped onto the base of the cage. By placing a resistance box in series with the solenoid and its power supply, the input voltage to the solenoid, and consequently the height of the pin, was varied, resulting in a graded and essentially linear series of outputs from the cage. This calibration system was used to determine both the linearity of cage output as a function of impact force as well as the overall sensitivity of each cage to vertical displacement. Consequently, each amplifier was adjusted so that all cages had equivalent outputs based on this calibration. To validate the impact calibration system, 10 rats were tested. On Day 1, 5 rats were placed in the MB-N504 systems and 5 were placed in the WPI system. Five minutes later the rats were presented with 50, 105-dB noise bursts presented at a 20-sec ISI. One day later, the same procedure was repeated except rats tested in the MB-N504 system were now tested in the WPI system and vice versa. The results were striking and unequivocal. Startle amplitudes measured in the MB-N504 system were within normal limits. However, startle amplitudes measured in the WPI system were consistently lower and outside our normal range. Relative to the MB-N504 system, the WPI system produced lower startle amplitudes, t(9)= 6.78, p <0.001, despite the observation that rats in the WPI system seemed to be startlingjust as much. It appeared that startle was only minimally detected in
FIG. 3. Fourier analysis of acoustic startle elicited with a 20 msec, 110 dB 4000 Hz tone measured with a wide band amplifier (low f'flter= 1 Hz, high f'flter=200 Hz). Panel A--normal rat startle. Panel B---startle 10 min after injection of 1.0 mg/kg strychnine. Panel C--normal startle in a cage containing a 11 kg lead brick to drastically change the resonant frequency of the cage. In all cases a dominant peak around 10 Hz occurs, which represents the frequency of startle.
this system, despite the fact that it had been calibrated with the solenoid to have output voltages equivalent to that of the MB-N504 system. Several experiments were conducted to determine the source of the inequity. After various interchanges of the cages, accelerometers and amplifiers it was concluded that startle amplitudes were systematically larger whenever the MB-N504 amplifier itself was used. To determine why the MB-N504 amplifier seemed better suited for measuring startle, the frequency-response function of this amplifier was assessed and compared to that of the WPI amplifier. This analysis showed that the MB-N504 amplifier was most sensitive to a 10 Hz signal and very insensitive to frequencies above about 50 Hz and below about 2 Hz. In contrast, the WPI amplifier had a fairly fiat response output from 2-100 Hz, and then fell off thereafter according to the f'dter installed to cut off high frequencies.
Frequency Analysis of the Rat's Startle Response The analysis of the MB-N504 amplifier suggested that a narrow pass f'dter centered about 10 Hz would be ideal for measuring startle. This led us to conclude that, contrary to the beliefs held by us and others in the field, the effective frequency of the rat startle response must be quite low, around 10 Hz. In fact, casual inspection of oscilloscope tracings of cage output resulting from a rat's startle (measured with a wide band amplifier) suggested a dominant frequency of about 10 Hz with a secondary one that is considerably higher at about 65 Hz (Fig. 2). This higher dominant frequency might represent the resonant frequency of the 860 g cage. The frequency of the rat's startle response was subsequently analyzed in a more systematic fashion. Auditory startle responses to 50 4-kHz tones presented at a 30-sec ISI were measured in the cage pictured in Fig. 1 using a fairly wide band amplifier (low frequency cutoff= 1 Hz, high frequency cutoff=4-kHz) and the Endevco 2217E accelerometer. Four animals were individually tested with 2 of these rats
380
CASSELLA AND DAVIS
5
/
2GQO
~e~ e
MB-N504 AMPLIFIER
/
2OOO
g3
n.UJ 1~00
13.
1000
I
/-
5OO
I'---" 20
30
40
G0
60
70
80
gO
100
2
~,~o._
.~_.o~ - * - ° -
d*. . . . *~- - ~ " * ~
/
1 p-
i
~
4
6
i --L_.
8 10
210
I
40
INPUT FREQUENCY
FREQUENCY
" - ~ -*
WPI-DAM5 AMPLIFIER
a ~
'~
100
I
200
(Hz.)
FIG. 4. Fourier analysis of cage output when an impact type calibrator was used. Calibrator consisted of a solenoid which dropped a 50 g pin from a height of 15 mm onto the cage floor. Note that there is no peak whatsoever at 10 Hz, indicating that an impact calibrator does not mimick rat startle in terms of the frequency of response.
FIG. 5. Frequency response of the MB-N504 amplifier and a WPIDAM5 amplifier fitted with a 200 Hz high frequency f'dter. The amplifiers were adjusted to give comparable outputs using the impact type calibrator which vibrated the cage primarily at a frequency of 55-60 Hz (at arrow). Despite equal sensitivity with this calibrator, the MB-N504 was much better for measuring startle, since it was much more sensitive at 10 Hz than the other amplifier.
injected with strychnine. This drug, which markedly increases startle [9], was used to determine if the frequency of high amplitude startle was comparable to the frequency of low amplitude startle. To drastically alter the resonant frequency of the cage, an 11 kg lead brick was placed in the cage and another rat was presented with 50 tones. Finally, 50 activations of the solenoid calibrator were given at a 30-sec interval. In all cases, the data were recorded on FM tape, and subsequently digitized and analyzed by Fourier transform using standard computer programs. Figure 3A shows a power spectrum for a normal rat startled in a cage like that shown in Fig. 1. A primary peak can be seen around 10--12 Hz with a secondary peak occurring at about 65 Hz. Figure 3B shows results for a rat given strychnine. Note again the primary frequency peak at 7-12 Hz and a secondary peak at 65-70 Hz. As previously suggested by casual inspection of oscilloscope tracing and indicated by Fourier analysis, the rat startle response produces two dominant sets of frequencies, one centered at about 10 Hz and the other centered at about 65 Hz. It is likely that one of these peaks represents the true frequency of the rat's startle while the other peak represents the resonant frequency of the cage when it is put into motion by the startling rat. It had been observed that when the cage is tapped with a metal rod and allowed to vibrate, it oscillates at about 55--65 Hz, suggesting that the lower frequency peak found with the power spectrum analysis represents the rat startle response. It was reasoned that the startle frequency would be unchanged when the resonant frequency of the cage was altered. The resonant frequency of the cage could be reduced by making the cage itself heavier. Therefore, the front and back walls of the cage were removed and an 11 kg lead brick was balanced on the floor bars. This changed the resonant frequency of the cage to about 20-25 Hz, as judged by tapping the cage with a metal rod. Figure 3C shows the Fourier analysis of a rat (placed on top of the brick) startling in this very heavy cage. Two dominant peaks were again found. One peak was now centered around 22 Hz, the new resonant frequency of the cage. This is dominant because the cage vibrates regularly and for a long time at this frequency since
it is no longer as effectively damped. Most important, however, is that even following this dramatic change in resonant frequency, a peak was still produced at about 10 Hz. Figure 4 shows the Fourier analysis of cage output when the solenoid calibrator was used. Two points are evident. First, the dominant frequency now peaks at about 56 Hz, the resonant frequency of the cage containing the relatively heavy calibrator. More importantly, however, is the absence of a peak at 10 Hz that is typically seen when a rat startles. Hence an impact-type calibrator basically moves the cage at its resonant frequency (plus some harmonics) but not at the frequency at which a rat startles.
Other Transducers To test the generality of the conclusion that startle consists of a 10 Hz perturbation of the cage, we substituted different transducers for the Endevco accelerometer. These included the original MB-Electronics accelerometer, a teflon coated magnet inserted halfway into a 28 volt relay coil, and a 14-cm midrange speaker. The accelerometer and the magnet in the coil were sandwiched between the bottom of the cage and the rubber stopper similar to the accelerometer pictured in Fig. 1. The speaker was placed next to the cage and cage movements were translated by a light metal rod rigidly mounted to the side of the cage in contact with the center of the speaker. Auditory startle responses elicited by 25 presentations of 110-dB noise bursts were recorded using the same rat placed in a cage equipped with each of the four different transducers, using a different transducer for each run of 25 auditory stimuli. In addition, 25 bangs of the impact calibrator were also presented, using each of the four transducers. The data were recorded on FM tape, digitized and analyzed by Fourier transform. For each of the four transducers, the rat vibrated the cage at a lower frequency than the impact calibrator did. Major pe~tks for the power spectrums obtained from rat startle responses were 5.0 (MB-Electronics accelerometer); 12.5 (Endevco accelerometer); 7.5 (magnet in the coil); and 7.5 (speaker). For the magnet in the coil a major peak also occurred at 60 Hz, probably reflecting the resonant frequency
S T A R T L E M E A S U R E M E N T SYSTEM
381
FIG. 6. Photograph of a startle cage calibrator consisting of a cut-down loud speaker that can be vibrated at 10 Hz. Lead weights sit inside the cone to provide extra mass. An accelerometer is bolted to the lead weights and is used to determine precisely the extent to which the cone moves.
of the cage as it connected to this transducer. Major peaks for the power spectrum obtained from the impact calibrator were 30 (Endevco accelerometer); 120 (MB-Electronics accelerometer); 30 (magnet in the coil); and 120 (speaker). Based on these data, we conclude that the rat startle response is roughly equivalent to a 5-15 Hz perturbation of the cage. While the exact value may vary somewhat depending on the cage and transducer that is used, the rat consistently vibrates the cage during startle at a lower frequency than that seen when the cage is set into motion by a mechanical impact. Hence an amplifier that preferentially measures a signal around 10 Hz should be optimal for measuring startle.
The Problem With Impact Type Calibrators A question that arises is why the WPI amplifier did not measure startle even though it was set to be equivalent to the MB-N504 amplifier according to the impact calibrator. Figure 5 shows the frequency characteristics of the MB-N504 amplifier and the WPI amplifier fitted with a 200 Hz high frequency fdter. When the impact calibrator was used (arrow) it basically vibrated the cage at its resonant frequency (55-65 Hz). The two amplifiers were thus set to give equivalent output voltages at about 55-65 Hz. Since the response characteristics of the WPI amplifier are essentially fiat at lower frequencies, its output at 10 H z would be about the same as that at 65 Hz. However, since the MB-N504 amplifier markedly attenuates signals above 10 Hz, its gain had to be substantially higher to give an equivalent output at 65 Hz. At such a gain, it was highly sensitive to a 10 Hz signal and therefore especially sensitive to startle. One could argue that comparable results would have been obtained if the gain o f the WPI amplifier had simply been turned up to give an ouptut comparable to the MB-N504 amplifier (i.e., if the dotted line in Fig. 5 were shifted up by
about 40 units). The problem with this argument, however, is that although startle would be measured adequately, so would all other cage movements, especially the resonant frequency of the cage. Hence any spontaneous, non-startle movements would be picked up. Much non-startle activity can be eliminated by measuring startle over a narrow time window after the eliciting stimulus. However, in some cases even this may not be adequate. F o r example, if a rat were given amphetamine (a drug that markedly increases activity as well as startle), the system might not discriminate between increases in startle and drug-induced increases in activity. In contrast, we have shown elsewhere that the MBN504 system does discriminate even when high doses of amphetamine are given [8]. Thus, the use of impact-type calibrators results in startle measurement systems that are not differentially sensitive to a frequency of response centered around the rat's startle.
A Valid Calibration System The data point to the conclusion that the startle response is like a 10 Hz signal. Therefore, the best way to calibrate startle cages would be to vibrate them at 10 Hz. To do this we took a large woofer (Radio Shack Cat No. 40:1275 A) and cut it down so that only the driver and cone remained. We then cut out lead circles, each weighing about 50 g and positioned them within the cone to add mass to the vibrating cone (Fig. 6). Care had to be taken to use parts that would not be attracted to the very strong driver magnet (e.g., we used lead weights and a stainless steel bolt to allow even stacking of the weights). A total weight of about 250 g worked b e s t - - a b o u t the size of a rat. This provided enough mass to easily move the cage, yet not so much that the speaker cone wabbled. A 10 Hz signal generated by a Wavetech signal generator (Model 182A) was used to drive
382
CASSELLA AND DAVIS
WPI-DAM5 SYSTEM
SYSTEM
MS-N504
100 9ol
/~re~
,/
STRYCHNINE
i~e~t~ jTRYCHNINE
I w _1
ao I 70 i
~ eo
JJ
\\
-J 5 0
~" 4 0
3°F.
~
w,..
10
{
1JO
3*0 410 1'0 2~0 3'0 MINUTES AFTER INJECTION
2~0
4~0
FIG. 7. Mean amplitude startle response over blocks of 6 stimuli (2-min periods) after injection of water or 1.0 mg/kg strychnine in the MB-N504 (left panel) or WPI-DAM5(right panel) startle test systems.
the cut-down speaker and, consequently, the cage at 10 Hz. An accelerometer (Endevco 2217 E) was mounted directly onto the lead weights to monitor the movement of the calibrator. In this way a constant, measurable frequency and amplitude of vibration was used to calibrate each cage by placing the calibrator inside the cage, setting the Wavetech to produce a 10 Hz vibration (as measured by the output of the accelerometer on the calibrator), and setting the cage accelerometer amplifier to produce a specific output. A Functional Startle Measurement System
To make the WPI amplifier comparable to the MB-N504, we added a 10 Hz band pass f'dter at the input stage of the amplifier. We then set all 10 cages (5 using MB-N504 amplifiers, 5 using the modified WPI amplifiers) to have
equivalent outputs according to the calibrator described above. The ultimate test of this new system was conducted using a total of 20 rats. On Day 1, 5 rats were injected with water and 5 with strychnine (1.0 mg/kg) and then presented with 90, 105 dB noise bursts at a 20-see ISI in the MB-N504 amplifier system. On Day 2, these procedures were repeated, except rats given water on Day 1 were now _given strychnine and vice versa. The other 10 rats were treated identically, except they were tested in the modified WPI amplifier system. On Days 3 and 4 these procedures were repeated once again but rats tested initially in the MB-N504 amplifier system were now tested in the modified WPI amplifier system and vice versa. In this way, all rats were tested in both amplifier systems after administration of water or strychnine, with the order of system testing and drug administration completely balanced. Figure 7 displays the results and two points are evident. First, baseline startle levels (startle after water) were essentially identical in both systems. Second, the magnitude and time course of strychnine's excitatory effect on startle were also essentially identical. An overall analysis of variance indicated a significant strychnine effect, F(1,19)=35.62, p<0.001, a significant time effect, F(19,361)=5.48,p<0.001, and a significant time × strychnine interaction, F(19,361)=7.60, p<0.001. However, there were no significant differences between the two startle amplifier systems nor any interactions involving these two startle systems. This is in marked contrast to the analysis of these systems prior to the addition of the 10 Hz filter to the WPI system. In conclusion, the rat startle response is best described as a 5-15 Hz signal. Hence transducers and amplifiers that are preferentially sensitive to about 10 Hz should be ideal for measuring startle while simultaneously filtering out cage movements produced by non-startle activity. In fact, expensive wide band amplifiers and accelerometers sensitive to high frequencies are not required at all. Lower cost accelerometers sensitive to low frequencies should be perfectly adequate. A calibration device that moves the startle cage at this frequency appears to be a valid method of standardizing the sensitivity of these units. ACKNOWLEDGEMENTS
This research was supported by NSF Grant BNS-81-20476, NIMH Grant MH-25642, Research Scientist Development Award MH-00004 to M.D. and the State of Connecticut. We thank Leslie Fields for typing the manuscript and Dr. John H. Kehne and Janice Mondlock for their helpful comments on this manuscript. Special thanks are extended to Greg McCarthy for performing the Fourier analysis and to Vaino Lipponen for critical consultation at many stages of this research.
REFERENCES 1. Brown, J. S., H. I. Kalish and I. E. Farber. Conditioned fear as revealed by magnitude of startle response to an auditory stimulus. J Exp Psychol 41: 317-327, 1951. 2. Carlton, P. L. and C. Advokat. Attenuated habituation due to parachlorophenylalanine. Pharmacol Biochem Behav 1: 657663, 1973. 3. Cunningham, C., C. R. Crowell, N. K. Eaton and J. S. Brown. A digital system for recording startle responses in small animals. Behav Res Methods lnstrum 5: 1-3, 1973. 4. Davis, M. Morphine and naloxone: Effects on conditioned fear as measured with the potentiated startle paradigm. Eur J Pharmacol 54: 341-347, 1979.
5. Davis, M. Diazepam and flurazepam: Effects on conditioned fear as measured with the potentiated startle paradigm. Psychopharmacology (Berlin) 62: 1-7, 1979. 6. Davis, M. Neurochemical modulation of sensory-motor reactivity: Acoustic and tactile startle reflexes. Neurosci Biobehav Rev 4: 241-263, 1980. 7. Davis, M. and D. 1. Astrachan. Conditioned fear and startle magnitude: Effects of different footshock or backshock intensifies used in training. J Exp Psychol (Animal Behav) 4: 95-103, 1978.
STARTLE MEASUREMENT
SYSTEM
8. Davis, M., T. H. Svensson and G. K. Aghajanian. Effects of dand I-amphetamine on habituation and sensitization of the acoustic startle response in rats. Psychopharmacologia 43: 1-11, 1975. 9. Kehne, J. H., D. W. Gallager and M. Davis. Strychnine: Brainstem and spinal mediation of excitatory effects on acoustic startle. Eur J Pharmacol 76: 177-186, 1981. 10. Hoffman, H. S. and M. Fleshier. An apparatus for the measurement of startle-response in the rat. A m J Psychol 74: 307309, 1964. 11. Hoffman, H. S. and J. R. Ison. Reflex modification in the domain of startle: I. Some empirical findings and their implications for how the nervous system processes sensory input. Psychol Rev 87: 175-189, 1980.
383 12. Leitner, D. S. and M. C. Rosenberger. A simple and inexpensive startle transducer with high output. Behav Res Methods lnstrum 15: 508--510, 1983. 13. Tischler, M. D. and M. Davis. A visual pathway that mediates fear-conditioned enhancement of acoustic startle. Brain Res 276: 55-72, 1983. 14. White, E. H. and M. Horlington. An apparatus for measuring startle response and motor activity in rats. Med Biol Eng 7: 325-327, 1969. 15. Wilson, C. J. and P. M. Groves. Refractory period and habituation of acoustic startle response in rats. J Comp Physiol Psychol 83: 492-498, 1973.
0031-9384/86 $3.00 + .00
BRIEF COMMUNICATION
The Design and Calibration of a Startle Measurement System J A M E S V. C A S S E L L A
AND MICHAEL
DAVIS
Yale University School o f Medicine, and the Ribicoff Research Facilities of the Connecticut Mental Health Center, 3 4 Park St., New Haven, CT 06508 R e c e i v e d 18 J u l y 1984 CASSELLA, J. V. AND M. DAVIS. The design and calibration of a startle measurement system. PHYSIOL BEHAV 36(2) 377-383, 1986.--The present study sought to determine appropriate instrumentation for amplification and calibration of cages used to measure the acoustic startle response in rats. Fourier analysis indicated that the characteristic frequency of the rat startle response is about 5-15 Hz. This value was consistent in cages differing widely in resonant frequency, among several different transducers and across a range of large and small startle responses. Given this characteristic frequency range of startle, it is suggested that amplifiers fitted with band pass filters centered at about 10 Hz should be ideal for measuring startle while simultaneously excluding non-startle activity. A device is described which vibrates startle cages at 10 Hz, since this seems most appropriate for calibrating the sensitivity of a startle system. Data are presented showing that this type of calibrator is more valid than an impact-type calibrator. Startle response
Amplifiers
Calibrators
T H E startle response in rodents is being used increasingly for the analysis of animal behavior. This reflex, which is typically elicited by a loud sound or air puff, has a short, reproducible latency and occurs in virtually every rat. Startle is a highly graded reflex that can be altered by changes in the parameters of the eliciting stimulus, the surrounding environmental stimuli, or general state of the animal [6,11]. During repetitive stimulation, startle habituates or sensitizes, depending on the exact parameters used, and has provided an excellent model system for the analysis of behavioral plasticity. Startle amplitude can be increased by prior fear conditioning [1,7] and currently is being used as a model system to evaluate the pharmacology [4,5] and anatomy [13] of classical conditioning. In addition, startle is sensitive to various drugs and is currently being used to investigate the pharmacology of behavior [6]. To date, many different systems have been described for measuring startle. In most cases, the startle response of the animal results in movement of a cage which is translated into a voltage, amplified, and then displayed, typically as a pen deflection or digital readout. A variety of transducers such as accelerometers [14], magnets within coils [10], photograph cartridges [2,3], strain gauges [15], piezo electric f'dm [12] or even purely mechanical systems [1] have been used. Recently, commercially designed startle cages have become available. A major, unsolved problem in the field of startle measurement is that no one has developed a satisfactory way to calibrate startle cages so that one cage is just as sensitive as another in measuring startle. This is both a problem within laboratories (since it would be desirable to have all cages in a
given laboratory set at a comparable sensitivity) as well as across laboratories (since it may be desirable to know how similar cage sensitivities are in different laboratories). Most calibration systems involve striking a startle cage with a reproducible force and measuring the resultant movement of the cage. F o r example, a solenoid may be positioned in a cage in such a way that when it is activated, the pin hits the cage which generates a voltage at impact. Dropping the pin from various heights or dropping pins of different weights can be used to determine the output linearity of the cage. In other cases, a solenoid is connected to a rigid superstructure. Activation of the solenoid moves an extension rod which hits the cage in a reproducible way [ 10]. Other workers have dropped clay or acrylic balls of different weights from various heights and recorded cage output [7,12]. While it is possible to make calibrators that automatically and reproducibly impact onto startle cages, it has never been clear if impact-type calibrators are valid devices for mimicking the behavior of the animal. In this paper, we will present evidence suggesting that impact-type calibrators may have little validity in this regard. Moreover, we will describe what is believed to be the essential aspect of the startle response dictating both how startle should be measured and how calibrators should be constructed.
The Problem of Amplification and Calibration of Startle Cages F o r the past several years we have been using the startle cage system shown in Fig. 1. Each of 5 stabilimeters consists
377
378
CASSELLA AND DAVIS
FIG. 1. Photograph of a startle cage. Accelerometer looks like a steel nut with a wire coming out of it and is sandwiched between bottom of the cage and a rubber stopper.
of an 8x 15× 15 cm Plexiglas and wire mesh cage suspended within a 25x20x20 cm steel frame. Within this frame the cage is rigidly sandwiched between 4 compression springs above, and a 5 x 5 cm rubber cylinder below. An accelerometer (M. B. Electronics Type 302), with a sensitivity of 40--60 mV/g, is located between the bottom of the cage and the top of the rubber cylinder. Its output is fed to an MB-N504 accelerometer amplifier. Cage movement results in a slight and imperceptible displacement of the accelerometer which generates a voltage that is proportional to the rate of cage displacement. Startle is defined as the peak accelerometer output that occurs within 200 msec after presentation of the startle-eliciting stimulus. Because the cage is rigidly held in place and no mechanical adjustments are required, the system has remained very stable over several years.
These 5 cages were calibrated by dropping a wooden arm attached to a rigid fulcrum onto the cage from a fixed height. By attaching different weights to the arm, graded outputs could be measured and the sensitivities of each accelerometer amplifier adjusted to give comparable outputs for all the cages. Increased research demands required construction of an additional 5 cages. A major change in this new system was to replace the M. B. Type 302 accelerometer, which was no longer available, with an Endevco Type 2217E accelerometer. The amplifier chosen for this new system was a WPIDAM 5 amplifier (WPI Instruments, Inc., New Haven, CT) since the one made by Endevco was too expensive. However, it was discovered that the WPI amplifier differed from the MB-N504 amplifier in that the WPI detected and
STARTLE M E A S U R E M E N T SYSTEM
379 I
A soo ACCELEROMETER
OUTPUT (z W 0
", -
I."
I
m'
t
STIMULUS
l,ovo,,
I0
20 30
40 50 60 70 80 90 100
2000
B 1500
20 MSEC.
BOO
MARKER
10 20 SO 40
50 eo 7o 80 oo 100
to
2o 3o 40 60 SO #o 8o go-~oo
FREQUENCY
FIG. 2. Oscilloscopic tracing of the output of the startle cage using a wide band amplifier (low frequency f'dter= 1 Hz, high frequency=4kHz). The rat was startled by a 20 msec, 4-kHz tone. The stimulus artifact in the beginningof the trace is present because the cage acts like a microphone and picks up the vibration produced by the tone, which is only partially eliminated by the high frequency f'dter. Within about 12 msec of stimulus onset, the cage begins to move down at a frequency of about 10 Hz. The higher, smaller amplitude frequency reflects the resonant frequency of the cage (about 55-60 Hz).
amplified the artifact produced by the 4-kHz tone. Consequently, a low pass f'flter was built into the WPI amplifier so that signals above about 200-Hz were f'dtered out, thus eliminating the stimulus artifact when the 4-kHz tone came on.
Having made these modifications to the WPI amplifier we attempted to calibrate the 5 cages using the MB-N504 amplifier and the 5 cages using the WPI amplifier so that all cages were equally sensitive to vertical displacement. The calibrator consisted of a 28 volt solenoid held rigidly in an aluminum frame that fit snugly into the startle cage. When 28 V was applied for 50 msec across the solenoid core, a 50 g pin was lifted to a height of 15 mm and then dropped onto the base of the cage. By placing a resistance box in series with the solenoid and its power supply, the input voltage to the solenoid, and consequently the height of the pin, was varied, resulting in a graded and essentially linear series of outputs from the cage. This calibration system was used to determine both the linearity of cage output as a function of impact force as well as the overall sensitivity of each cage to vertical displacement. Consequently, each amplifier was adjusted so that all cages had equivalent outputs based on this calibration. To validate the impact calibration system, 10 rats were tested. On Day 1, 5 rats were placed in the MB-N504 systems and 5 were placed in the WPI system. Five minutes later the rats were presented with 50, 105-dB noise bursts presented at a 20-sec ISI. One day later, the same procedure was repeated except rats tested in the MB-N504 system were now tested in the WPI system and vice versa. The results were striking and unequivocal. Startle amplitudes measured in the MB-N504 system were within normal limits. However, startle amplitudes measured in the WPI system were consistently lower and outside our normal range. Relative to the MB-N504 system, the WPI system produced lower startle amplitudes, t(9)= 6.78, p <0.001, despite the observation that rats in the WPI system seemed to be startlingjust as much. It appeared that startle was only minimally detected in
FIG. 3. Fourier analysis of acoustic startle elicited with a 20 msec, 110 dB 4000 Hz tone measured with a wide band amplifier (low f'flter= 1 Hz, high f'flter=200 Hz). Panel A--normal rat startle. Panel B---startle 10 min after injection of 1.0 mg/kg strychnine. Panel C--normal startle in a cage containing a 11 kg lead brick to drastically change the resonant frequency of the cage. In all cases a dominant peak around 10 Hz occurs, which represents the frequency of startle.
this system, despite the fact that it had been calibrated with the solenoid to have output voltages equivalent to that of the MB-N504 system. Several experiments were conducted to determine the source of the inequity. After various interchanges of the cages, accelerometers and amplifiers it was concluded that startle amplitudes were systematically larger whenever the MB-N504 amplifier itself was used. To determine why the MB-N504 amplifier seemed better suited for measuring startle, the frequency-response function of this amplifier was assessed and compared to that of the WPI amplifier. This analysis showed that the MB-N504 amplifier was most sensitive to a 10 Hz signal and very insensitive to frequencies above about 50 Hz and below about 2 Hz. In contrast, the WPI amplifier had a fairly fiat response output from 2-100 Hz, and then fell off thereafter according to the f'dter installed to cut off high frequencies.
Frequency Analysis of the Rat's Startle Response The analysis of the MB-N504 amplifier suggested that a narrow pass f'dter centered about 10 Hz would be ideal for measuring startle. This led us to conclude that, contrary to the beliefs held by us and others in the field, the effective frequency of the rat startle response must be quite low, around 10 Hz. In fact, casual inspection of oscilloscope tracings of cage output resulting from a rat's startle (measured with a wide band amplifier) suggested a dominant frequency of about 10 Hz with a secondary one that is considerably higher at about 65 Hz (Fig. 2). This higher dominant frequency might represent the resonant frequency of the 860 g cage. The frequency of the rat's startle response was subsequently analyzed in a more systematic fashion. Auditory startle responses to 50 4-kHz tones presented at a 30-sec ISI were measured in the cage pictured in Fig. 1 using a fairly wide band amplifier (low frequency cutoff= 1 Hz, high frequency cutoff=4-kHz) and the Endevco 2217E accelerometer. Four animals were individually tested with 2 of these rats
380
CASSELLA AND DAVIS
5
/
2GQO
~e~ e
MB-N504 AMPLIFIER
/
2OOO
g3
n.UJ 1~00
13.
1000
I
/-
5OO
I'---" 20
30
40
G0
60
70
80
gO
100
2
~,~o._
.~_.o~ - * - ° -
d*. . . . *~- - ~ " * ~
/
1 p-
i
~
4
6
i --L_.
8 10
210
I
40
INPUT FREQUENCY
FREQUENCY
" - ~ -*
WPI-DAM5 AMPLIFIER
a ~
'~
100
I
200
(Hz.)
FIG. 4. Fourier analysis of cage output when an impact type calibrator was used. Calibrator consisted of a solenoid which dropped a 50 g pin from a height of 15 mm onto the cage floor. Note that there is no peak whatsoever at 10 Hz, indicating that an impact calibrator does not mimick rat startle in terms of the frequency of response.
FIG. 5. Frequency response of the MB-N504 amplifier and a WPIDAM5 amplifier fitted with a 200 Hz high frequency f'dter. The amplifiers were adjusted to give comparable outputs using the impact type calibrator which vibrated the cage primarily at a frequency of 55-60 Hz (at arrow). Despite equal sensitivity with this calibrator, the MB-N504 was much better for measuring startle, since it was much more sensitive at 10 Hz than the other amplifier.
injected with strychnine. This drug, which markedly increases startle [9], was used to determine if the frequency of high amplitude startle was comparable to the frequency of low amplitude startle. To drastically alter the resonant frequency of the cage, an 11 kg lead brick was placed in the cage and another rat was presented with 50 tones. Finally, 50 activations of the solenoid calibrator were given at a 30-sec interval. In all cases, the data were recorded on FM tape, and subsequently digitized and analyzed by Fourier transform using standard computer programs. Figure 3A shows a power spectrum for a normal rat startled in a cage like that shown in Fig. 1. A primary peak can be seen around 10--12 Hz with a secondary peak occurring at about 65 Hz. Figure 3B shows results for a rat given strychnine. Note again the primary frequency peak at 7-12 Hz and a secondary peak at 65-70 Hz. As previously suggested by casual inspection of oscilloscope tracing and indicated by Fourier analysis, the rat startle response produces two dominant sets of frequencies, one centered at about 10 Hz and the other centered at about 65 Hz. It is likely that one of these peaks represents the true frequency of the rat's startle while the other peak represents the resonant frequency of the cage when it is put into motion by the startling rat. It had been observed that when the cage is tapped with a metal rod and allowed to vibrate, it oscillates at about 55--65 Hz, suggesting that the lower frequency peak found with the power spectrum analysis represents the rat startle response. It was reasoned that the startle frequency would be unchanged when the resonant frequency of the cage was altered. The resonant frequency of the cage could be reduced by making the cage itself heavier. Therefore, the front and back walls of the cage were removed and an 11 kg lead brick was balanced on the floor bars. This changed the resonant frequency of the cage to about 20-25 Hz, as judged by tapping the cage with a metal rod. Figure 3C shows the Fourier analysis of a rat (placed on top of the brick) startling in this very heavy cage. Two dominant peaks were again found. One peak was now centered around 22 Hz, the new resonant frequency of the cage. This is dominant because the cage vibrates regularly and for a long time at this frequency since
it is no longer as effectively damped. Most important, however, is that even following this dramatic change in resonant frequency, a peak was still produced at about 10 Hz. Figure 4 shows the Fourier analysis of cage output when the solenoid calibrator was used. Two points are evident. First, the dominant frequency now peaks at about 56 Hz, the resonant frequency of the cage containing the relatively heavy calibrator. More importantly, however, is the absence of a peak at 10 Hz that is typically seen when a rat startles. Hence an impact-type calibrator basically moves the cage at its resonant frequency (plus some harmonics) but not at the frequency at which a rat startles.
Other Transducers To test the generality of the conclusion that startle consists of a 10 Hz perturbation of the cage, we substituted different transducers for the Endevco accelerometer. These included the original MB-Electronics accelerometer, a teflon coated magnet inserted halfway into a 28 volt relay coil, and a 14-cm midrange speaker. The accelerometer and the magnet in the coil were sandwiched between the bottom of the cage and the rubber stopper similar to the accelerometer pictured in Fig. 1. The speaker was placed next to the cage and cage movements were translated by a light metal rod rigidly mounted to the side of the cage in contact with the center of the speaker. Auditory startle responses elicited by 25 presentations of 110-dB noise bursts were recorded using the same rat placed in a cage equipped with each of the four different transducers, using a different transducer for each run of 25 auditory stimuli. In addition, 25 bangs of the impact calibrator were also presented, using each of the four transducers. The data were recorded on FM tape, digitized and analyzed by Fourier transform. For each of the four transducers, the rat vibrated the cage at a lower frequency than the impact calibrator did. Major pe~tks for the power spectrums obtained from rat startle responses were 5.0 (MB-Electronics accelerometer); 12.5 (Endevco accelerometer); 7.5 (magnet in the coil); and 7.5 (speaker). For the magnet in the coil a major peak also occurred at 60 Hz, probably reflecting the resonant frequency
S T A R T L E M E A S U R E M E N T SYSTEM
381
FIG. 6. Photograph of a startle cage calibrator consisting of a cut-down loud speaker that can be vibrated at 10 Hz. Lead weights sit inside the cone to provide extra mass. An accelerometer is bolted to the lead weights and is used to determine precisely the extent to which the cone moves.
of the cage as it connected to this transducer. Major peaks for the power spectrum obtained from the impact calibrator were 30 (Endevco accelerometer); 120 (MB-Electronics accelerometer); 30 (magnet in the coil); and 120 (speaker). Based on these data, we conclude that the rat startle response is roughly equivalent to a 5-15 Hz perturbation of the cage. While the exact value may vary somewhat depending on the cage and transducer that is used, the rat consistently vibrates the cage during startle at a lower frequency than that seen when the cage is set into motion by a mechanical impact. Hence an amplifier that preferentially measures a signal around 10 Hz should be optimal for measuring startle.
The Problem With Impact Type Calibrators A question that arises is why the WPI amplifier did not measure startle even though it was set to be equivalent to the MB-N504 amplifier according to the impact calibrator. Figure 5 shows the frequency characteristics of the MB-N504 amplifier and the WPI amplifier fitted with a 200 Hz high frequency fdter. When the impact calibrator was used (arrow) it basically vibrated the cage at its resonant frequency (55-65 Hz). The two amplifiers were thus set to give equivalent output voltages at about 55-65 Hz. Since the response characteristics of the WPI amplifier are essentially fiat at lower frequencies, its output at 10 H z would be about the same as that at 65 Hz. However, since the MB-N504 amplifier markedly attenuates signals above 10 Hz, its gain had to be substantially higher to give an equivalent output at 65 Hz. At such a gain, it was highly sensitive to a 10 Hz signal and therefore especially sensitive to startle. One could argue that comparable results would have been obtained if the gain o f the WPI amplifier had simply been turned up to give an ouptut comparable to the MB-N504 amplifier (i.e., if the dotted line in Fig. 5 were shifted up by
about 40 units). The problem with this argument, however, is that although startle would be measured adequately, so would all other cage movements, especially the resonant frequency of the cage. Hence any spontaneous, non-startle movements would be picked up. Much non-startle activity can be eliminated by measuring startle over a narrow time window after the eliciting stimulus. However, in some cases even this may not be adequate. F o r example, if a rat were given amphetamine (a drug that markedly increases activity as well as startle), the system might not discriminate between increases in startle and drug-induced increases in activity. In contrast, we have shown elsewhere that the MBN504 system does discriminate even when high doses of amphetamine are given [8]. Thus, the use of impact-type calibrators results in startle measurement systems that are not differentially sensitive to a frequency of response centered around the rat's startle.
A Valid Calibration System The data point to the conclusion that the startle response is like a 10 Hz signal. Therefore, the best way to calibrate startle cages would be to vibrate them at 10 Hz. To do this we took a large woofer (Radio Shack Cat No. 40:1275 A) and cut it down so that only the driver and cone remained. We then cut out lead circles, each weighing about 50 g and positioned them within the cone to add mass to the vibrating cone (Fig. 6). Care had to be taken to use parts that would not be attracted to the very strong driver magnet (e.g., we used lead weights and a stainless steel bolt to allow even stacking of the weights). A total weight of about 250 g worked b e s t - - a b o u t the size of a rat. This provided enough mass to easily move the cage, yet not so much that the speaker cone wabbled. A 10 Hz signal generated by a Wavetech signal generator (Model 182A) was used to drive
382
CASSELLA AND DAVIS
WPI-DAM5 SYSTEM
SYSTEM
MS-N504
100 9ol
/~re~
,/
STRYCHNINE
i~e~t~ jTRYCHNINE
I w _1
ao I 70 i
~ eo
JJ
\\
-J 5 0
~" 4 0
3°F.
~
w,..
10
{
1JO
3*0 410 1'0 2~0 3'0 MINUTES AFTER INJECTION
2~0
4~0
FIG. 7. Mean amplitude startle response over blocks of 6 stimuli (2-min periods) after injection of water or 1.0 mg/kg strychnine in the MB-N504 (left panel) or WPI-DAM5(right panel) startle test systems.
the cut-down speaker and, consequently, the cage at 10 Hz. An accelerometer (Endevco 2217 E) was mounted directly onto the lead weights to monitor the movement of the calibrator. In this way a constant, measurable frequency and amplitude of vibration was used to calibrate each cage by placing the calibrator inside the cage, setting the Wavetech to produce a 10 Hz vibration (as measured by the output of the accelerometer on the calibrator), and setting the cage accelerometer amplifier to produce a specific output. A Functional Startle Measurement System
To make the WPI amplifier comparable to the MB-N504, we added a 10 Hz band pass f'dter at the input stage of the amplifier. We then set all 10 cages (5 using MB-N504 amplifiers, 5 using the modified WPI amplifiers) to have
equivalent outputs according to the calibrator described above. The ultimate test of this new system was conducted using a total of 20 rats. On Day 1, 5 rats were injected with water and 5 with strychnine (1.0 mg/kg) and then presented with 90, 105 dB noise bursts at a 20-see ISI in the MB-N504 amplifier system. On Day 2, these procedures were repeated, except rats given water on Day 1 were now _given strychnine and vice versa. The other 10 rats were treated identically, except they were tested in the modified WPI amplifier system. On Days 3 and 4 these procedures were repeated once again but rats tested initially in the MB-N504 amplifier system were now tested in the modified WPI amplifier system and vice versa. In this way, all rats were tested in both amplifier systems after administration of water or strychnine, with the order of system testing and drug administration completely balanced. Figure 7 displays the results and two points are evident. First, baseline startle levels (startle after water) were essentially identical in both systems. Second, the magnitude and time course of strychnine's excitatory effect on startle were also essentially identical. An overall analysis of variance indicated a significant strychnine effect, F(1,19)=35.62, p<0.001, a significant time effect, F(19,361)=5.48,p<0.001, and a significant time × strychnine interaction, F(19,361)=7.60, p<0.001. However, there were no significant differences between the two startle amplifier systems nor any interactions involving these two startle systems. This is in marked contrast to the analysis of these systems prior to the addition of the 10 Hz filter to the WPI system. In conclusion, the rat startle response is best described as a 5-15 Hz signal. Hence transducers and amplifiers that are preferentially sensitive to about 10 Hz should be ideal for measuring startle while simultaneously filtering out cage movements produced by non-startle activity. In fact, expensive wide band amplifiers and accelerometers sensitive to high frequencies are not required at all. Lower cost accelerometers sensitive to low frequencies should be perfectly adequate. A calibration device that moves the startle cage at this frequency appears to be a valid method of standardizing the sensitivity of these units. ACKNOWLEDGEMENTS
This research was supported by NSF Grant BNS-81-20476, NIMH Grant MH-25642, Research Scientist Development Award MH-00004 to M.D. and the State of Connecticut. We thank Leslie Fields for typing the manuscript and Dr. John H. Kehne and Janice Mondlock for their helpful comments on this manuscript. Special thanks are extended to Greg McCarthy for performing the Fourier analysis and to Vaino Lipponen for critical consultation at many stages of this research.
REFERENCES 1. Brown, J. S., H. I. Kalish and I. E. Farber. Conditioned fear as revealed by magnitude of startle response to an auditory stimulus. J Exp Psychol 41: 317-327, 1951. 2. Carlton, P. L. and C. Advokat. Attenuated habituation due to parachlorophenylalanine. Pharmacol Biochem Behav 1: 657663, 1973. 3. Cunningham, C., C. R. Crowell, N. K. Eaton and J. S. Brown. A digital system for recording startle responses in small animals. Behav Res Methods lnstrum 5: 1-3, 1973. 4. Davis, M. Morphine and naloxone: Effects on conditioned fear as measured with the potentiated startle paradigm. Eur J Pharmacol 54: 341-347, 1979.
5. Davis, M. Diazepam and flurazepam: Effects on conditioned fear as measured with the potentiated startle paradigm. Psychopharmacology (Berlin) 62: 1-7, 1979. 6. Davis, M. Neurochemical modulation of sensory-motor reactivity: Acoustic and tactile startle reflexes. Neurosci Biobehav Rev 4: 241-263, 1980. 7. Davis, M. and D. 1. Astrachan. Conditioned fear and startle magnitude: Effects of different footshock or backshock intensifies used in training. J Exp Psychol (Animal Behav) 4: 95-103, 1978.
STARTLE MEASUREMENT
SYSTEM
8. Davis, M., T. H. Svensson and G. K. Aghajanian. Effects of dand I-amphetamine on habituation and sensitization of the acoustic startle response in rats. Psychopharmacologia 43: 1-11, 1975. 9. Kehne, J. H., D. W. Gallager and M. Davis. Strychnine: Brainstem and spinal mediation of excitatory effects on acoustic startle. Eur J Pharmacol 76: 177-186, 1981. 10. Hoffman, H. S. and M. Fleshier. An apparatus for the measurement of startle-response in the rat. A m J Psychol 74: 307309, 1964. 11. Hoffman, H. S. and J. R. Ison. Reflex modification in the domain of startle: I. Some empirical findings and their implications for how the nervous system processes sensory input. Psychol Rev 87: 175-189, 1980.
383 12. Leitner, D. S. and M. C. Rosenberger. A simple and inexpensive startle transducer with high output. Behav Res Methods lnstrum 15: 508--510, 1983. 13. Tischler, M. D. and M. Davis. A visual pathway that mediates fear-conditioned enhancement of acoustic startle. Brain Res 276: 55-72, 1983. 14. White, E. H. and M. Horlington. An apparatus for measuring startle response and motor activity in rats. Med Biol Eng 7: 325-327, 1969. 15. Wilson, C. J. and P. M. Groves. Refractory period and habituation of acoustic startle response in rats. J Comp Physiol Psychol 83: 492-498, 1973.