Methodological optimization of tinnitus assessment using prepulse inhibition of the acoustic startle reflex

BR A IN RE S E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

Available online at www.sciencedirect.com

www.elsevier.com/locate/brainres

Research Report

Methodological optimization of tinnitus assessment using prepulse inhibition of the acoustic startle reflex R.J. Longenecker, A.V. Galazyuk⁎ Department of Anatomy and Neurobiology, Northeast Ohio Medical University, Rootstown, OH, USA

A R T I C LE I N FO

AB S T R A C T

Article history:

Recently prepulse inhibition of the acoustic startle reflex (ASR) became a popular technique for

Accepted 25 February 2012

tinnitus assessment in laboratory animals. This method confers a significant advantage over

Available online 4 March 2012

the previously used time-consuming behavioral approaches utilizing basic mechanisms of conditioning. Although this technique has been successfully used to assess tinnitus in differ-

Keywords:

ent laboratory animals, many of the finer details of this methodology have not been described

Prepulse inhibition of the acoustic

enough to be replicated, but are critical for tinnitus assessment. Here we provide detail de-

startle reflex

scription of key procedures and methodological issues that provide guidance for newcomers

Gap detection

with the process of learning to correctly apply gap detection techniques for tinnitus assess-

Tinnitus

ment in laboratory animals. The major categories of these issues include: refinement of hardware for best performance, optimization of stimulus parameters, behavioral considerations, and identification of optimal strategies for data analysis. This article is part of a Special Issue entitled: Tinnitus Neuroscience. © 2012 Published by Elsevier B.V.

1.

Introduction

For several decades pre-pulse inhibition of the startle reflex has been successfully used as a powerful tool to identify various psychiatric disorders in humans and the deficiencies associated with sensorimotor gating in laboratory animals (Braff et al., 2001). The basic parameters of the stimulus paradigm used for this testing have been well studied. The stimulus parameters have been optimized for accurate assessment (Carlson and Willott, 1996; Hoffman and Searle, 1965, Ison et al., 2002, 2005). The neural circuits that relate to this phenomenon have also been identified and intensively studied (Koch and Schnitzler, 1997). Recently prepulse inhibition of the acoustic startle reflex (ASR) has been adapted and successfully tested as a powerful technique for tinnitus assessment in laboratory animals

(Turner et al., 2006). This method confers a significant advantage over the previously used time-consuming behavioral approaches utilizing basic mechanisms of conditioning (Bauer et al., 1999; Guitton et al. 2003; Heffner and Harrington, 2002; Heffner and Koay 2005; Lobarinas et al. 2004; Rüttiger et al., 2003). It does not require animal training. Tinnitus assessment can be done in animals within a single short testing session. This method relies on a reduction of the acoustic startle reflex by a preceding silent gap in an otherwise constant acoustic background. Animals with behavioral evidence of tinnitus cannot detect silence and therefore their reduction of the startle reflex is significantly less than in normal animals. This method has been successfully used to assess tinnitus induced by salicylate overdose or acoustic trauma in rats (Kraus et al. 2010; Turner et al., 2006; Wang et al., 2009; Yang et al., 2007; Zhang et al., 2011) and mice (Longenecker and Galazyuk, 2011;

⁎ Corresponding author at: Northeast Ohio Medical University, Department of Anatomy and Neurobiology, 4209 State Route 44, Rootstown, OH 44272, USA. Fax: + 1 330 325 5916. E-mail address: [email protected] (A.V. Galazyuk). 0006-8993/$ – see front matter © 2012 Published by Elsevier B.V. doi:10.1016/j.brainres.2012.02.067

BR A I N R ES E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

55

Middleton et al., 2011). Many of the finer details of this methodology, however, have not been described enough to be replicated, but are critical for tinnitus assessment. All these details can be roughly divided into four major categories: refinement of hardware for best performance, optimization of stimulus parameters, behavioral considerations, and identification of optimal strategies for data analysis. Thus, the purpose of this paper is to help newcomers with the typically painful process of learning to correctly apply gap detection techniques for tinnitus assessment in laboratory animals.

2.

Results

2.1.

Hardware refinement

Any lab equipment requires tuning for the best performance including those designed to measure the behavioral response of an animal to sensory stimuli. Here we focus on refining a system aimed to assess gap detection performance. This method has recently become popular for tinnitus assessment in laboratory animals. Currently, such systems are commercially available and several labs have designed their own systems. All these systems share the same principals but vary slightly in design. Here we want to emphasize the most general points of system refinement, which we found to be critical for better system performance.

2.1.1.

Addressing an issue of speakers' nonlinearity

Depending on the design, prepulse inhibition testing stations (boxes) have either one (Sun et al., 2009) or two loud speakers (Turner et al., 2006). In a two speaker system one speaker presents a startle stimulus while another presents either a continuous background interrupted by a gap of silence or a prepulse. Typically, speakers at a given input have nonlinear frequency transfer functions. As a result, such a speaker may have very dissimilar (sometimes more than 15 dB) sound intensities at different frequencies. Fig. 1A shows frequency transfer functions of seven Fostex FT17H speakers calibrated with a 1/4-inch microphone (Brüel and Kjaer 4135). It is important to make adjustments (typically via software) in order to make all frequency outputs roughly equal in amplitude. Such an adjustment is much more critical for a speaker presenting a narrow band noise centered at different frequencies than for a wide band noise startle speaker. Some systems might have the ability to adjust speakers at different testing stations independently. If this option is available, it should be used. However, if your system consists of multiple testing stations but only one correction can be used, you may face a problem. This problem arises from the fact that each speaker may have a different frequency transfer function even though each speaker is the same make and model (Fig. 1A). The question then becomes how to make all speakers' frequency transfer function linear by using just one correction file. The solution is very simple: a set of speakers for all testing stations (possibly even a few extra, for future possible repairs) should be purchased from the same manufacturer with a particular emphasis that the speakers are made during the same manufacturing cycle time (the same batch). Typically, these speakers have very similar frequency transfer functions

Fig. 1 – Frequency/amplitude output functions for 7 loudspeakers. Dotted line represents the target calibrated amplitude (75 dB) for all frequencies tested. A) Loudspeakers which were made at different manufacturing cycles (batches). B) Loudspeakers which were made at the same manufacturing cycle. C) Loudspeakers shown in B after software adjustment for speaker nonlinearity using one correction file applied to all speakers.

(Fig. 1B). Installation of these speakers will allow for one correction file to adjust the frequency/amplitude output in a linear manner (Fig. 1C). In this case, any given background frequency for the gap detection paradigm of all simultaneously tested animals, in different testing stations, will receive equal sound intensity. This will allow comparisons of results collected from different animals and different testing stations.

2.1.2.

Restrainer issues

Animal restrainers come in all shapes and sizes. The design of the restrainer is critical in several respects to how an animal will respond to sound stimuli. First, it is important that the space between the restrainer and the animal should be minimal in an effort to minimize undesired movements during the startle stimulus. The animal should not be able to move freely and should not be able to rear up (Turner et al., 2006; Walton et al., 1997). On the other hand, the animal should not be hindered completely

56

BR A IN RE S E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

because this will likely lead to undue stress during testing. Adjustment of restrainers to meet these requirements is important for the success of ASR testing. Second, it has been shown in many studies that the duration (Ison et al., 2005), depth (Ison et al., 1998), and rise/fall time of the gap significantly contribute to its salience to an animal (Ison et al., 2002). We found that restrainers with unnecessarily large reverberating walls may drastically alter sound parameters and reduce animal performance in gap detection. We evaluated gap parameters inside two types of mouse restrainer for Kinder Scientific Ltd. which manufactures systems for behavioral assessment of tinnitus in laboratory animals using prepulse inhibition of the ASR. An ultrasonic condenser microphone (CM16/CMPA, Avisoft Bioacoustics) was placed inside these restrainers. The output from the microphone was fed to a gain-adjusted amplifier and A/D converter (UltraSoundGate 416H, Avisoft Bioacoustics). Acoustic signals were sampled at 250 kHz with 16-bit depth and monitored in real time with Avisoft software (Version 5.1.01, Avisoft Bioacoustics). In Fig. 2, an example of an acoustically inferior mouse restrainer is shown. This restrainer was not easily permeable to sounds due to large solid plastic walls (Fig. 2A). Most importantly, this restrainer made the gap onset longer which made this gap undetectable for mice (Fig. 2B). We speculate that the mechanism underlying large wall effects

A

on the gap onset was as follows. Continuous narrow band noise where the gap was induced may have made the large wall of the restrainer reverberate. This reverberation continued for a few milliseconds after a gap of silence was induced, which made the gap onset very shallow. Another tested restrainer was easily permeable to sound presumably because it did not have large plastic walls (Fig. 3A). The gap onset in this restrainer was much sharper (Fig. 3B). All mice tested in this restrainer showed very consistent and robust gap detection performance. Thus, acoustic properties of the animal restrainer are vital for gap detection experiments and for behavioral assessment of tinnitus in laboratory animals.

2.1.3.

Echo suppression

Most outside and inside walls of startle testing stations are made of hard laminated materials which may cause some undesirable sound reflections or echoes. Such reflections can alter the quality of sound during testing and more importantly the quality of the gap during gap detection testing. Such effects can be greatly reduced by adding echo suppressive material to the inside walls of the testing stations. In our lab we use Sonex foam from Pinta Acoustics. This foam improved the sound quality and made the gap fall time even more distinct (compare the gap quality in Figs. 4B and 3B). One thing to keep in mind is that after adding foam, the overall sound levels will be attenuated by about 10 dB.

2.2.

Optimization of stimulus parameters

2.2.1.

Startle and background sound levels

Prepulse inhibition of the acoustic startle reflex is a wellestablished phenomenon. Multiple studies have determined the basic parameters used to induce robust and reliable inhi-

A

B

B

20 ms Fig. 2 – A sound-unfriendly mouse restrainer alters onset of the gap. A) A restrainer with large reverberating walls. B) A waveform of continuous narrow band noise centered at 10 kHz with 20 ms gap of silence recorded inside the restrainer. Note long lasting gap onset.

20 ms Fig. 3 – A sound-friendly mouse restrainer makes onset of the gap sharper. A) A small restrainer permeable to sound. B) A waveform of narrow band noise with 20 ms gap of silence recorded inside the restrainer. Note an improved gap onset compared to the gap shown in Fig. 2B.

57

BR A I N R ES E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

A

A

Startle stimulus

STARTLE trial

Narrow band noise

B Gap of silence

GAP+STARTLE trial

B 100 ms

20 ms Fig. 4 – A startle testing station with sound-friendly mouse restrainer and with echo suppressive foam installed on inside walls. A) A view of the testing station. B) A waveform of narrow band noise with 20 ms gap of silence recorded inside the restrainer. Note near ideal shape of the gap with sharp onset.

bition of the startle response when a gap of silence, in an otherwise continuous narrowband background noise, centered at different frequencies, precedes the startle stimulus (Hoffman and Searle, 1965; Koch and Schnitzler, 1997; Willott et al., 1994; Yang et al., 2007). Tinnitus is assessed by comparing an animal's startle response to a single broadband noise burst (STARTLE) presented in otherwise continuous narrowband background noise with the startle response when STARTLE was preceded by a gap of silence (GAP + STARTLE) (Fig. 5). Ratios are calculated by dividing the GAP + STARTLE responses by STARTLE responses. A ratio of 1 means that the animal does not detect a gap, whereas ratio values lower than 1 indicates better detection. The typical ratios for control (unexposed) animals range between 0.7 and 0.5 which represents a gap induced suppression of the startle between 30 and 50%, respectively. The sound exposed mice at a given frequency typically exhibit gap detection deficits at or near the frequencies of exposure because their tinnitus fills the gap of silence. This method is the most sensitive when both the background frequency and intensity match the animal's tinnitus because the salience of the gap would be diminished maximally under these conditions (Bauer et al., 1999). Unfortunately, we do not know the tinnitus intensity for a given animal. Therefore, only the center frequency of the narrowband noise is varied, whereas sound intensities are kept constant at the level in a range of 60 to 75 dB SPL assuming that the animal's tinnitus approximates this intensity (Turner et al., 2006). The startle stimulus intensity is typically set at 100 to 120 dB SPL in most ASR paradigms (Heffner and Heffner, 2001).

Fig. 5 – Two types of stimuli for assessing gap detection performance in mice.Gap detection stimulus paradigm consists of A) STARTLE trial — a startle stimulus of wide band noise (20 ms duration, 110 dB SPL) embedded in a continuous background narrow band noise centered at 10, 12.5, 16, 20, 25, and 31.5 kHz and presented at 75 dB SPL; B) GAP + STARTLE trial — similar to the STARTLE trial with the addition of a 20 ms gap of silence embedded 100 ms before the startle stimulus.

We also found that these ranges of background and startle intensities are appropriate for assessing gap detection deficits in mice following sound exposure (Longenecker and Galazyuk, 2011).

2.2.2.

Masking effects of background on startle

When developing a stimulus session for prepulse inhibition or gap detection tests, it is important to be aware of the masking effects of the background on the startle stimulus (Carlson and Willott, 2001). If not addressed appropriately, this suppression may lead to inappropriate conclusions concerning tinnitus identification. It has been shown that a continuous background suppresses an animal's response to startle stimuli. This suppression largely depends on both the background frequency and intensity (Gerrard and Ison, 1990). As demonstrated in Fig. 6A an individual mouse has shown a “U” shape pattern of frequency dependent suppression of a startle response. We found this suppression in all tested mice without exception. The suppression also increases with increasing background intensity (Fig. 6B). We found that the shape of this curve approximates the audiogram of the CBA/CaJ mouse (Radziwon et al., 2009) and several other rodents (Heffner et al., 2001). Such similarity strongly suggests that at the frequencies where mice have the minimal response threshold (12.5–16 kHz) the startle suppression was maximal. To the best of our knowledge, the similarity between these two curves has not been reported before. We hypothesize that this is a universal phenomenon in the animal kingdom. This hypothesis needs further validation.

58

BR A IN RE S E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

Fig. 6 – Startle response amplitude in CBA/CaJ mice depends on both center frequency and intensity of continuous narrow band noise. A) Startle response amplitude as a function of center frequency of continuous narrow band noise presented at 70 dB SPL and recorded three times from the same mouse within a one week period. B) Startle response amplitude as a function of narrow band noise frequency recorded from the same mouse at two noise intensities (50 and 70 dB SPL).

Since the startle suppression by the background is very similar for both the trials with or without a gap at a given background frequency, the ratio between the two measurements should be frequency independent. However, if this suppression is too strong the startle amplitude will be minimally affected by typical gap suppression (Fig. 7, outlined data point). If so, the gap/no gap ratio for this frequency would have a value close to 1, which could lead to an inappropriate conclusion that no gap detection is present and more importantly that this represents the presence of tinnitus at this frequency. To avoid this issue, during data analysis of a gap detection test, it is vital to compare the level of background animal movement when no sound is presented with the level of startle response for each individual frequency when no gap was introduced. In our experiments we determine the mean and standard deviation values for all trials when no sound was introduced. Then, if the mean value of startle (no gap) trials at a given frequency is equal to or less than the mean plus the standard deviation of background movement, the ratio should not be calculated for this frequency (Fig. 7, outlined by a circle). If this problem occurs, the startle stimulus intensity should be increased and the testing session should be repeated.

2.2.3.

function raised steeply until it was saturated at the level of about 110 dB SPL. When the mice were tested with the gap detection paradigm at different startle intensities (105 vs. 115 dB SPL) they showed very different gap detection performance. This performance was very poor when startle stimuli were presented at the level of saturation, at 115 dB SPL (Fig. 8B).

Choosing an appropriate startle intensity

As was mentioned previously, ASR studies in the past have employed a range of startle amplitudes. Each species and strain of animals has its own startle stimulus–response function. Such response function curves have been shown for many strains of mice (Bullock et al., 1997). These curves also were obtained in our laboratory from CBA/CaJ mice (Fig. 8A). The startle stimuli were presented at the levels ranged from 50 to 130 dB SPL in a random fashion. The interstimulus interval was randomized within a range between 7 and 15 s. Each startle intensity was repeated six times. In average our mice exhibited startle threshold at the level of about 80 dB SPL. At higher startle intensities, the startle stimulus–response

Fig. 7 – If frequency/intensity-dependent suppression of the startle response is not considered, it may lead to an incorrect conclusion regarding tinnitus assessment with the gap detection paradigm. Narrow band noise presented at 70 dB SPL suppresses startle response amplitude at 12.5 kHz center frequency (outlined data point) to a level which is not significantly different from animal movements without any stimulus presented. Horizontal dashed line indicates mean value of animal's movement without any stimulus with the standard deviation represented by a shaded bar.

BR A I N R ES E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

59

Fig. 8 – Startle stimulus intensity is critical for gap detection performance. A) Startle stimulus/response function recorded from 7 CBA/CaJ mice. The thick black line represents the average function over 7 mice. The dashed horizontal lines indicate the level of startle response saturation and the level 25% below saturation. B, C) Gap detection performance in the same mice measured at 115 and 105 dB SPL startle stimulus intensities, respectively. Dashed horizontal lines indicate the ratio of 1 that represents poor of no gap detection performance, whereas the rations less than 1 a better gap detection.

However, the gap detection performance was greatly improved when startle stimuli were at about 25%–50% below the level of saturation, for instance 105 dB SPL (Fig. 8C). This data strongly suggest that avoiding saturated startle response levels is critical when your goal is to alter the startle response with a preceded gap. Thus, before tinnitus assessment using the gap detection paradigm, the startle stimulus/response function should be obtained. Then, using this function the startle stimuli should be set at the levels falling within the rising phase.

2.2.4.

typically enough for the exposed mice to improve their performance in the gap detection test. One more phenomenon needs to be mentioned here: gap induced facilitation of startle responses. A gap preceded by a startle stimulus occasionally facilitates startle responses. This phenomenon has been previously mentioned in the ASR literature (Ison and Hammond, 1971; Ison et al., 1998). At present there is not a clear understanding of the underlying mechanism of this facilitation. However, it is very likely that such facilitation indicates gap detection similar to that indi-

Considering effects of sound exposure

In our recent work we found that sound exposure suppresses startle responses even when sound exposure is performed monaurally (Longenecker and Galazyuk, 2011). Such changes in startle responses may potentially complicate gap detection tests. However, we found that even a suppressed startle after a mild sound exposure (115–120 dB SPL for 1 h) is strong enough to be inhibited by a preceding gap. Nevertheless some minor adjustments should be done to improve the effectiveness of gap detection tests on sound exposed animals. Similar to unexposed mice, the sound exposed mice exhibited a “U” shape pattern of frequency dependent suppression of a startle response, although this function was less pronounced (Fig. 9). The startle stimulus/response function in these animals was also different compared to the unexposed group (compare curves in Figs. 8A and 10). This function rose much slower than in the unexposed group (0.0025 N/dB and 0.0075 N/dB, respectively). Sound exposed mice often required 5 dB to 10 dB higher startle stimulus intensity in order to reach the startle response level that was 75% of the maximum startle response amplitude (Fig. 10). Such an adjustment was

Fig. 9 – Startle response amplitude in 4 CBA/CaJ mice after sound exposure as a function of continuous narrow band noise presented at different center frequency and at the level of 70 dB SPL.

60

BR A IN RE S E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

amplitude typically decreases as a session progress with the rate about 1.5%/min. Therefore, a single testing session should not last very long. Otherwise the startle amplitude will decrease to the point when it will have no room to be inhibited by a preceded gap. However, more than one testing session (usually 2–3) can be conducted on the same mouse in the same day if animals would be given 30 to 40 min rest between sessions.

2.3.2.

Fig. 10 – Startle stimulus/response function recorded from 4 CBA/CaJ mice after sound exposure. The thick black line represents the average function over 4 mice. The dashed horizontal lines indicate the level of startle response saturation and the level 25% below saturation.

cated by gap induced inhibition because in both cases the presence of the gap changes the response to the startle. Further investigation is necessary to shed light on mechanisms underlying the gap induced facilitation.

2.3.

Behavioral considerations

2.3.1.

Habituation issues

Habituation is a phenomenon which is present in nearly every behavioral task, and has been well documented (Davis, 1974). Habituation occurs because “a response eliciting stimulus itself fails to predict any biologically important event and, hence, is no longer behaviorally relevant (Koch, 1999)”. The startle response amplitude, threshold, and latency are modulated by habituation within a testing session (Pilz and Schnitzler, 1996). Habituation has also been observed in prepulse inhibition of the acoustic startle reflex (Lipp and Krinitzky, 1997). To avoid such habituation the inter-trial interval (ITI) must be an appropriate duration and needs to be randomized (Davis, 1970). If the ITI is too short, the animal may habituate to the startle very quickly. In mice, for instance, we noticed that ITIs randomized within a range between 7 and 15 s work well. This ensures that the animal cannot predict the next startle stimulus. We also found that stimulus habituation can be further suppressed if each trial type is uncorrelated with a particular ITI. In other words, an animal should not predict that a particular trial type will always be linked to a certain ITI. Even after all these precautions we found that some animals, especially older ones, exhibit habituation to a gap detection session lasting more than 40 min (Fig. 11). In our experiments the startle response

Acclimation

Although not as well studied, acclimation is yet another important phenomenon to account for. It was shown that startle responses change significantly between the first couple testing sessions (Faraday and Grunberg, 2000). The solution to this would be to expose naive animals to stimulus sessions several times before starting the study. We found that after three to five sessions the behavior usually stabilized and thus the data could be consistently analyzed. Minor changes in stimulus presentation can also change the behavioral response. If any parameter is changed, a new acclimation exposure should be given before collecting data. This will insure the animal will be properly adjusted to the paradigm and that your data will be reliable.

2.3.3.

Multiple sessions

Behavioral data is inherently filled with variables that can strongly shape results. Some of these variables can be controlled as discussed above, but it is near impossible to control for all possible behavioral contexts. For this reason, animals should be tested several times under a similar condition to obtain a decent picture of their behavioral responses. Multiple testing times can provide options for analyzing data. Should data from multiple sessions representing several days of work be averaged or should the best data set of a single session be used? This is a question that will be left up to interpretation. It should be noted that having those options, as a result of testing an animal multiple times in the same relative condition, will give you viable options for analyzing data. This would not be the case if each animal is tested just once at a given time or condition.

Fig. 11 – Habituation of gap detection performance in 7 nine-month-old CBA/CaJ mice. Dashed line indicates gap/no poor gap detection performance.

BR A I N R ES E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

2.3.4.

Animal stress

Nearly every action taken in a lab can cause an animal stress (Balcombe et al., 2004). Stress is known to impair cognitive function (Arnsten, 2009), and thus it can alter the results of behavioral tests (Dawood et al., 2004; Graham et al., 2011; Kaneto, 1995). It has been shown that loud unexpected sounds can raise levels of stress-related hormones (Burow et al., 2005). Both restraint and social stress are common in laboratory animals (Stone and Quartermain, 1997). Simple handling of an animal, putting it in a new environment, and cage changes can also raise levels of stress in animals (Duke et al., 2001; Seggie and Brown, 1974). Some visible indications of animal stress include increased bowl movements (observed after a session), intense resistance to being picked up, and unwillingness to enter the restrainer before testing. In our study we also noticed that the most “clean” data were obtained from the mice which exhibited very little fecal boli accumulation in the restrainer after a testing session. Perhaps more intuitively, any paradigm in which an animal is repeatedly startled would elicit a stress related response (Götz and Janik, 2011). For this reason it is important to: (1) minimize the amount of time the animal spends in the testing stations, (2) avoid testing animals during the day their cages were cleaned, (3) transport animals from animal facilities to the testing place with minimal stress (cart with soft wheels, animals' cages are covered), (4) avoid exposing animals to loud devices such as electric sweepers or audio equipment, (5) avoid handling the animal more than necessary.

2.4.

Data analysis

The intensity of pre-pulse inhibition of a startle reflex has often been expressed as a ratio. This ratio is calculated by taking the response amplitude of a startle preceded by a prepulse (or gap in otherwise continuous background) and dividing by the response amplitude when a startle is presented alone. Typically the startle response magnitude in an individual animal can vary from one presentation to another. Therefore, each stimulus needs to be presented multiple times (typically 10 to 15) in order to obtain a sufficient measurement of animal's behavioral response to this stimulus. As in any behavioral paradigm an animal's responses will fluctuate. For this reason it is common for an animal to occasionally exhibit either an unusually strong startle response or to show no startle at all. Such outliers can significantly alter even averaged data over multiple presentations. To address this issue many researchers simply remove the largest and the smallest data points from each data set collected by multiple presentation of the same stimulus (Jeskey and Willott, 2000; Turner et al., 2006). Instead, we propose to use Grubb's statistical test for identification of outliers (Grubbs, 1969; Stefansky, 1972). This test detects one outlier at a time. This outlier should be extracted from the dataset and the test should be repeated until no additional outliers are detected. This test should not be used for sample sizes of six or less. Based on our experience this test typically does not identify more than one (very rarely two) outliers in one data set containing 15 data points.

61

If outliers are identified and removed, the number of gap trials may not correspond to the number of no-gap trials at a given background frequency. In this situation how should the ratio and standard error for a given background frequency be calculated? We propose the following solution. First, a mean of no-gap startle responses should be calculated. Then, each data point for gap trials at this frequency should be divided by the no-gap mean. This will yield as many ratios as there are gap trials after outliers are removed. The mean and standard error from these ratios should be calculated. We found that such data treatment resulted in consistent measurements of the gap detection performance over multiple testing sessions collected from an individual animal.

Acknowledgments We thank Dr. Merri Rosen for valuable comments on earlier versions of this manuscript and Marie Gadziola for technical assistance with acoustical assessments of mouse restrainers. The authors also thank Olga Galazyuk for developing software that allowed off-line data analysis and statistical evaluation. This study is supported by the Tinnitus Research Consortium (AVG).

REFERENCES

Arnsten, A.F.T., 2009. Stress signaling pathways that impair prefrontal cortex structure and function. Nat. Rev. Neurosci. 10, 410–422. Balcombe, J.P., Bernard, N.D., Sandusky, C., 2004. Laboratory routines cause animal stress. Contemp. Top. Lab. Anim. Sci. 43, 42–51. Bauer, C.A., Brozoski, T.J., Rojas, R., Boley, J., Wyder, M., 1999. Behavioral model of chronic tinnitus in rats. Otolaryngol. Head Neck Surg. 121, 457–462. Braff, D.L., Geyer, M.A., Swerdlow, N.R., 2001. Human studies of prepulse inhibition of startle: normal subjects, patient groups, and pharmacological studies. Psychopharmacology 156, 234–258. Bullock, A.E., Slobe, B.S., Vazquez, V., Collins, A.C., 1997. Inbred mouse strains differ in the regulation of startle and prepulse inhibition of the startle response. Behav. Neurosci. 111, 1353–1360. Burow, A., Day, H.E.W., Campeau, S., 2005. A detailed characterization of loud noise stress: intensity analysis of hypothalamo-pituitary-adrenocortical axis and brain activation. Brain Res. 1062, 63–73. Carlson, S., Willott, J.F., 1996. The behavioral salience of tones as indicated by prepulse inhibition of the startle response: relationship to hearing loss and central neural plasticity in C57BL/6J mice. Hear. Res. 99, 168–175. Carlson, S., Willott, J.F., 2001. Modulation of the acoustic startle response by background sound in C57BL/6J mice. In: Willott, James F. (Ed.), Handbook of Mouse Auditory Research: from Behavior to Molecular Biology. CRC Press, Boca Raton, pp. 83–90. Davis, M., 1970. Effects of interstimulus interval length and variability on startle-response habituation in the rat. J. Comp Physiol. Psychol. 72, 177–192. Davis, M., 1974. Signal-to-noise ratio as a predictor of startle amplitude and habituation in the rat. J. Comp Physiol. Psychol. 86, 812–825. Dawood, M.Y., Lumley, L.A., Robison, C.L., Saviolakis, G.A., Meyerhoff, J.L., 2004. Accelerated Barnes maze test in mice for assessment of stress effects on memory. Ann. N. Y. Acad. Sci. 1032, 304–307.

62

BR A IN RE S E A RCH 1 4 85 ( 20 1 2 ) 5 4 –62

Duke, J.L., Zammit, T.G., Lawson, D.M., 2001. The effects of routine cage-changing on cardiovascular and behavioral parameters in male Sprague–Dawley rats. Contemp. Top. Lab. Anim. Sci. 40, 17–20. Faraday, M.M., Grunberg, N.E., 2000. The importance of acclimation in acoustic startle amplitude and pre-pulse inhibition testing of male and female rats. Pharmacol. Biochem. Behav. 66, 375–381. Gerrard, R.L., Ison, J.R., 1990. Spectral frequency and the modulation of the acoustic startle reflex by background noise. J. Exp. Psychol. Anim. Behav. Process. 16, 106–112. Götz, T., Janik, V.M., 2011. Repeated elicitation of the acoustic startle reflex leads to sensitisation in subsequent avoidance behaviour and induces fear conditioning. BMC Neurosci. 12 (30), 1–12. Graham, L.K., Yoon, T., Kim, J.J., 2011. Stress impairs optimal behavior in a water foraging choice task in rats. Learn. Mem. 17, 1–4. Grubbs, F., 1969. Procedures for detecting outlying observations in samples. Technometrics 11 (1), 1–21. Guitton, M.J., Caston, J., Ruel, J., Johnson, R.M., Pujol, R., Puel, J.L., 2003. Salicylate induces tinnitus through activation of cochlear NMDA receptors. J Neurosci. 23, 3944–3952. Heffner, H.E., Harrington, I.A., 2002. Tinnitus in hamsters following exposure to intense sound. Hear Res. 170, 83–95. Heffner, H.E., Heffner, R.S., 2001. Behavioral assessment of hearing in mice. In: Willott, James F. (Ed.), Handbook of Mouse Auditory Research: from Behavior to Molecular Biology. CRC Press, Boca Raton, pp. 19–29. Heffner, H.E., Koay, G., 2005. Tinnitus and hearing loss in hamsters (Mesocricetus auratus) exposed to loud sound. Behav. Neurosci. 119, 734–742. Heffner, R.S., Koay, G., Heffner, H.E., 2001. Audiograms of five species of rodents: implications for the evolution of hearing and the perception of pitch. Hear. Res. 157, 138–152. Hoffman, H.S., Searle, J.L., 1965. Acoustic variables in the modification of startle reaction in the rat. J. Comp. Physiol. Psychol. 60, 53–58. Ison, J.R., Hammond, G.R., 1971. Modification of the startle reflex in the rat by changes in the auditory and visual enviroments. J. Comp. Physiol. Psychol. 75, 435–452. Ison, J.R., Agrawal, P., Pak, J., Vaughn, W.J., 1998. Changes in temporal acuity with age and with hearing impairment in the mouse: a study of the acoustic startle reflex and its inhibition by brief decrements in noise level. J. Acoust. Soc. Am. 104, 1696–1704. Ison, J.R., Castro, J., Allen, P., Virag, T.M., Walton, J.P., 2002. The relative detectability for mice of gaps having different ramp durations at their onset and offset boundaries. J. Acoust. Soc. Am. 112, 740–747. Ison, J.R., Allen, P.D., Rivoli, P.J., Moore, J.T., 2005. The behavioral response of mice to gaps in noise depends on its spectral components and its bandwidth. J. Acoust. Soc. Am. 117, 3944–3951. Jeskey, J.E., Willott, J.F., 2000. Modulation of prepulse inhibition by an augmented acoustic environment in DBA/2J mice. Behav. Neurosci. 114, 991–997. Kaneto, H., 1995. Learning/memory processes under stress conditions. Behav. Brain Res. 83, 71–74. Koch, M., 1999. The neurobiology of startle. Prog. Neurobiol. 59, 107–128. Koch, M., Schnitzler, H., 1997. The acoustic startle response in rats — circuits mediating evocation, inhibition and potentiation. Behav. Brain Res. 89, 35–49.

Kraus, K.S., Mitra, S., Jimenez, Z., Hinduja, S., Ding, D., Jiang, H., Gray, L., Lobarinas, E., Sun, W., Salvi, R.J., 2010. Noise trauma impairs neurogenesis in the rat hippocampus. Neuroscience 167, 1216–1226. Lipp, O.V., Krinitzky, S.P., 1997. The effect of repeated prepulse and reflex stimulus presentations on startle prepulse inhibition. Biol. Psychol. 47, 67–76. Lobarinas, E., Sun, W., Cushing, R., Salvi, R., 2004. A novel behavioral paradigm for assessing tinnitus using schedule-induced polydipsia avoidance conditioning (SIP-AC). Hear Res. 190, 109–114. Longenecker, R.J., Galazyuk, A.V., 2011. Development of tinnitus in CBA/CaJ mice following sound exposure. J. Assoc. Res. Otolaryngol. 12, 647–658. Middleton, J.W., Kiritani, T., Pedersen, C., Turner, J.G., Shepherd, G.M., Tzounopoulos, T., 2011. Mice with behavioral evidence of tinnitus exhibit dorsal cochlear nucleus hyperactivity because of decreased GABAergic inhibition. Proc. Natl. Acad. Sci. U. S. A. 108, 7601–7606. Pilz, P.K.D., Schnitzler, H., 1996. Habituation and sensitization of the acoustic startle response in rats: amplitude, threshold, and latency measures. Neurobiol. Learn. Mem. 66, 67–79. Radziwon, K.E., June, K.M., Stolzberg, D.J., Xu-Friedman, M., Salvi, R.J., Dent, M.L., 2009. Behaviorally measured audiograms and gap detection thresholds in CBA/CaJ mice. J. Comp. Physiol. A 195, 1–19. Rüttiger, L., Ciuffani, J., Zenner, H.P., Knipper, M., 2003. A behavioral paradigm to judge acute sodium salicylate-induced sound experience in rats: a new approach for an animal model on tinnitus. Hear Res. 180, 39–50. Seggie, J.A., Brown, G.M., 1974. Stress response patterns of plasma corticosterone, prolactin, and growth hormone in the rat, following handling or exposure to novel environment. Can. J. Physiol. Pharmacol. 53, 629–637. Stefansky, W., 1972. Rejecting outliers in factorial designs. Technometrics 14, 469–479. Stone, E.A., Quartermain, D., 1997. Greater behavioral effects of stress in immature as compared to mature male mice. Physiol. Behav. 63, 143–145. Sun, W., Lu, J., Stolzberg, D., Gray, L., Deng, A., Lobarinas, E., Salvi, R.J., 2009. Salicylate increases the gain of the central auditory system. Neuroscience 159, 325–334. Turner, J.G., Brozoski, T.J., Bauer, C.A., Parrish, J.L., Myers, K., Hughes, L.F., Caspary, D.M., 2006. Gap detection deficits in rats with tinnitus: a potential novel screen tool. Behav. Neurosci. 120, 188–195. Walton, J.P., Frisina, R.D., Ison, J.R., O'Neill, W.E., 1997. Neural correlates of behavioral gap detection in the inferior colliculus of the young CBA mouse. J. Comp. Physiol. A 181, 161–176. Wang, H., Brozoski, T.J., Turner, J.G., Ling, L., Parrish, J.L., Hughes, L.F., Caspary, D.M., 2009. Plasticity at glycinergic synapses in dorsal cochlear nucleus of rats with behavioral evidence of tinnitus. Neuroscience 164, 747–759. Willott, J.F., Carlson, S., Chen, H., 1994. Prepulse inhibition of the startle response in mice: relationship to hearing loss and auditory system plasticity. Behav. Neurosci. 108, 703–713. Yang, G., Lobarinas, E., Zhang, L., Turner, J., Stolzberg, D., Salvi, R., Sun, W., 2007. Salicylate induced tinnitus: behavioral measures and neural activity in auditory cortex of awake rats. Hear. Res. 226, 244–253. Zhang, J., Zhang, Y., Zhang, X., 2011. Auditory cortex electrical stimulation suppresses tinnitus in rats. J. Assoc. Res. Otolaryngol. 12, 185–201.