| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Psychology, University of California, Berkeley, California 94720; 2 Department of Psychology, University of Washington, Seattle, Washington 98195
Submitted 2 July 2003; accepted in final form 23 September 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
; Dooling and Searcy 1980; Eales 1987; Marler 1970; Marler and Peters 1977; ten Cate 1991). This preference is evident even in acoustically naïve birds (Braaten and Reynolds 1999) and is often referred to as the "innate auditory preference" (see Marler 1997). Because the behavioral acoustic preferences that guide song learning appear to exist independent of experience, such preferences must to some extent reflect the inherent acoustic tuning properties of neurons in the songbird auditory system.
The locus of neurons that contribute to conspecific song preferences is unknown. Although neurons in the song control system of the forebrain respond selectively to the bird's own song, these neurons do not respond well to conspecific song (Doupe and Konishi 1991; Margoliash 1983). There are several reasons to indicate that the auditory midbrain region, mesencephalicus lateralis, dorsalis (MLd) is a candidate locus for tuning to conspecific sounds. 1) The MLd is analogous to the mammalian inferior colliculus (IC) and the amphibian torus semicircularis (TS), where specialized acoustic tuning properties related to vocal signals have been demonstrated. Studies on vocalizing frogs, echolocating bats, and mice indicate that complex acoustic tuning properties matching the acoustic features of conspecific vocalizations arise in the auditory midbrains of these animals (Casseday and Covey 1992; Casseday et al. 1997; Diekamp and Gerhardt 1992; Feng et al. 1978; Given 1993; Mittmann and Wenstrup 1995; Narins and Capranica 1980; Pollak and Bodenhamer 1981; Rose and Capranica 1983, 1985; Suga 1969). 2) Virtually all of the multiple, parallel auditory brain stem pathways converge in the MLd, suggesting this as a major site of neural integration where convergent inputs could interact to produce signal transformations necessary for specialized tuning (Fig. 1; Akesson et al. 1987; Boord 1968; Conlee and Parks 1986; Correia et al. 1982; Karten 1967, 1968; Park and Pollak 1994; Parks and Rubel 1975; Takahashi and Konishi 1988; Wild 1995). 3) The songbird MLd may be hypertrophied (Cobb 1964; Rylander 1979). 4) The songbird MLd receives descending input from regions that are involved in the control of song behavior, providing a potential feedback loop for the integration of auditory information and vocal motor control (Fig. 1; Mello et al. 1998; Wild 1993).
|
) recently showed that neurons in the ascending auditory forebrain region, field L (primary auditory cortex analog), respond preferentially to song over synthetic sounds with matched power spectra. Thus it seems possible that acoustic tuning in the ascending auditory nuclei could contribute to the mechanisms of auditory preference for conspecific vocalizations. To begin to understand the role of the songbird auditory midbrain in the processing of acoustic information, we studied the basic auditory responses of single neurons in the zebra finch MLd. This is, to our knowledge, the first examination of the response properties of auditory midbrain neurons in a songbird. We describe here the tonotopy, temporal response patterns, frequency tuning, intensity coding, and latency relationships of MLd neurons in adult male zebra finches.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
We used 23 adult male zebra finches that were purchased from a supplier (Magnolia Bird Farm, Anaheim, CA). Birds were housed in groups of 5 to 10 individuals and maintained on a 14:10 light/dark cycle in a temperature-controlled room. Food and water were available at all times. All animal procedures were approved by the University of Washington Animal Care Committee.
Surgical preparation
Birds were anesthetized with urethane (Sigma, St. Louis, MO; 2.5 mg/g, given in 4 intramuscular injections delivered at 20-min intervals). A bird was placed on a platform attached to a stereotaxic head holder. Body temperature was continuously monitored and adjusted to between 38 and 39°C using a custom-designed heater with a thermistor placed in the cloaca and a heating blanket placed under the bird. Lidocaine was applied to the skin overlying the skull region covering the dorsal midline of the brain and the optic lobe. After lidocaine application, a small incision was made in the scalp along the midline of the cranium and a metal post was fixed to the surface of the skull using dental acrylic. Another incision was made in the skin over the skull covering the optic tectum, caudal to the eye and dorsal to the ear. A small opening was made in the skull overlying the optic tectum and the dura was resected from the surface of the brain.
Sound generation and stimuli
Sound stimuli were generated using custom software and a digital signal processor [DSP; Tucker Davis Technologies (TDT)]. The output of the DSP was routed through a digital to analog converter (TDT), through an anti-aliasing filter (TDT), then through digital attenuators (TDT), and finally through an analog attenuator and audio amplifier (Krohn-Hite 7500) to a free-field loudspeaker (custom) that was placed 15 in. in front of the bird's head. The output of the loudspeaker was measured at the beginning of each experiment with a 
-in. Larson-Davis microphone.
The stimuli were pure tones (0.1–7.0 kHz) of durations between 5 and 1,000 ms, presented at sound levels between 5 and 90 dB, depending on the test. Stimuli always started at the zero-crossing point of the sine wave. Rise–fall times varied across stimuli but were generally 2 ms. Stimuli were presented either singly, as pairs separated by a variable delay, or as sequences of multiple sounds. Presentation rates varied between 1 and 3 repetitions/s. Within a trial, stimuli were presented in a random sequence. Each stimulus was presented 15 to 25 times.
Recordings
All recordings were made inside a walk-in double-walled sound-attenuation booth (Industrial Acoustics). The activity of single neurons was recorded extracellularly using glass micropipettes filled with 1 M NaCl. In some experiments, electrodes also contained 5% biotinylated dextran amine (BDA) to mark electrode locations by iontophoretic injection. Pipette tip diameters were typically <1.0 µm and impedances (at 1 kHz) ranged from 5 to 25 M
. Electrodes were aimed using visual landmarks and advanced in 1.0-µm steps using a Kopf hydraulic microdrive. Data were collected only from units that could be identified as cell bodies with reasonable certainty, in that they had a signal-to-noise ratio of 
3:1 and biphasic action potential waveform. Recordings were amplified with a negative capacitance electrometer, filtered (300 Hz high pass, 10 kHz low pass), displayed on a Tektronix 5113 multichannel oscilloscope and monitored on an audio amplifier and loudspeaker. Spikes were discriminated with a TDT spike discriminator. Spike times were collected in 1-µs bins by TDT event timers and stored on a computer using custom software. Dot raster displays, spike rates, spike latencies, total spike number, peristimulus time histograms (PSTHs), and excitatory frequency response areas (FRAs) were seen on-line. More PSTHs and other graphic displays of the data were generated off-line. Using a combination of commercial and custom software packages, further statistical analyses such as interspike interval histograms, tuning curves, rate-level functions (RLFs), means and variability of spike latencies, and rate-duration functions were performed.
Search stimuli
White noise was used as a search stimulus while approaching MLd. Once significant multiunit activity was detected, a variety of search stimuli including pure tones, noise bands, white noise, FM sweeps, and bird calls were used to isolate single units. A wide variety of stimuli were used to reduce the likelihood that neurons were isolated in a biased fashion based on the sounds presented. This variation in search stimuli is particularly important in the zebra finch MLd because we found that almost no cells in this region show spontaneous firing under urethane anesthesia.
Data analysis
To examine temporal response patterns, frequency coding, and intensity coding, complete excitatory FRAs were generated. FRA plots show a grid of PSTHs, one for each frequency/intensity combination across a range of frequencies and intensities. For every unit, an FRA was generated by recording the responses to 50-ms pure tones ranging in frequency between 0.5 and 7.0 kHz (within the frequency range of the zebra finch audiogram) and ranging in intensity between below threshold and 90 dB. Frequency steps were 0.5 kHz for intensity ranges above threshold and 0.1 kHz around threshold. Intensity steps were 5 dB at intensities above threshold and 2 dB around threshold. Frequencies were presented in random order. For most cells, multiple FRAs and additional tests such as RLFs were used to define spontaneous firing rate, response pattern (e.g., onset, primary-like, etc.), characteristic frequency (CF), threshold, width of frequency tuning, and first-spike latencies. Spontaneous activity rates were calculated from the 50 to 100 ms collection windows preceding each stimulus repetition. Threshold was defined as the lowest intensity to elicit a response on 10 out of 15 trials. CF was defined as the frequency with the lowest threshold. Frequency tuning was examined by calculating tuning curve bandwidths at 80 dB SPL and Q10 and Q30. Q10 is equal to the CF divided by the linear bandwidth measured at 10 dB above threshold; Q30 is the same but uses the bandwidth measured at 30 dB SPL. Examples of FRAs are shown in Fig. 4.
|
Histology
Electrode locations were histologically verified using iontophoretic injections of BDA (5-µA DC current pulsed 7 s ON/7 s OFF). Usually, one to 2 recording sites within a pass were marked. After recording sessions, birds were overdosed with Nembutal (0.5 mg/g body weight) and transcardially perfused using formalin (10%). Brains were immediately dissected free of the head and postfixed for 2 days in formalin. After postfixation, brains were cryoprotected in 30% sucrose in distilled water for 24 h and embedded in gelatin with formalin for 3–4 days. Frontal sections (40 µm) were cut on a freezing microtome and collected in 2 series into PBS. One series was mounted on gelatin subbed slides and stained for Nissl with cresyl violet. The second series was incubated briefly in Triton X-100 in PBS and then in avidin-biotin complex solution (ABC Vectastain Elite Kit, Vector Laboratories). After PBS rinses, BDA injection sites were visualized by incubation in diaminobenzidine (DAB) and 0.001% hydrogen peroxide in PBS. Sections were then rinsed in PBS, mounted onto chrome-alum slides, dried, and coverslipped with DPX.
Localization of auditory neurons in MLd/tonotopy
The borders between MLd and the surrounding nucleus intercollicularis (ICo), and the subregions of MLd have not been systematically defined in the zebra finch as they have in the barn owl (see Knudsen 1999). Therefore we used BDA injections (Fig. 2A) and Nissl-stained sections (Fig. 2B) to determine the borders of the area responsive to sound. Generally, two injections were made to label a pass, usually at the borders of the region responsive to sound. We then compared the locations of the injections with publications and brain atlases describing the MLd/ICo complex in numerous avian species including zebra finch to localize our recordings sites. The medial, dorsal, and ventral limits of MLd were often visible in Nissl sections. BDA injections that were placed to mark the medial and lateral borders of the auditory area matched the borders we identified using Nissl-stained sections. In the most rostral and most caudal portions of the MLd/ICo complex, the lateral borders of MLd were not visible in Nissl-stained sections. Because of this uncertainty, it is possible that some of the more rostral and lateral recording sites extended into ICo. As indicated by BDA injections labeling the borders of auditory responsive areas, the most lateral region from which we recorded was generally responsive to acoustic stimuli, and single units from this region were included in our analysis.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Figure 2A shows the location of a BDA injection within MLd. Figure 2B shows the adjacent Nissl-stained section. The CFs of cells systematically increased as electrode passes moved from dorsal to ventral positions. Figure 2C illustrates the anatomical locations and CFs of cells near the dorsal and ventral borders of MLd. The recording sites shown are from different birds. The recording locations were estimated in each bird using BDA injections and then projected onto a schematic drawing representing MLd. The tonotopic pattern showing an increase in CF in the dorsal to ventral orientation is the same as that found in the mammalian IC. No clear arrangement of temporal response patterns was found.
Spontaneous activity
Most (100 out of 111) MLd units showed no spontaneous activity (Fig. 3). Eleven of 111 cells showed low spontaneous activity (<7 spikes/s). No specific CF was associated with the presence of spontaneous firing, but each cell with some spontaneous firing showed either a "primary-like" or "sustained" response to tones (see following text).
|
Because we are ultimately interested in how cells in MLd process acoustic communication signals, we examined frequency and intensity coding together by collecting excitatory FRA plots (see above). Figure 4 shows examples of FRAs from 2 different cells to illustrate temporal response patterns and frequency tuning. Each cell fired with a consistent temporal response pattern across frequencies and intensities. One cell had an ongoing or sustained response pattern and was narrowly tuned to frequency (Fig. 4A). The other cell had an onset pattern and was broadly tuned to frequency (Fig. 4B).
Temporal response patterns
FRA plots were used to determine temporal response patterns for each cell. Additional responses to tones at CF between threshold and 90 dB SPL were used to confirm response pattern. Of 111 cells, 92 showed clear response patterns. The 19 cells for which a response pattern could not be assigned either had responses that were too variable to characterize or changed response patterns when intensity changed. Those cells with consistent responses were first classified as responding transiently with only 1–4 spikes per stimulus (onset) or responding throughout the duration of the stimulus (ongoing). Of the cells showing ongoing responses, 3 distinct temporal response patterns were observed. These 3 response patterns correspond to the "primary-like," "sustained," and "primary-like with notch" response patterns observed in single neurons of both mammalian and avian auditory nuclei (Koppl and Carr 2003; Pollak et al. 1978; Syka et al. 2000; Warchol and Dallos 1990). All consistent responses fell into 4 categories: 1) onset, in which a few spikes were locked to the onset of the stimulus; 2) primary-like, in which the response showed strong firing at the onset of the stimulus followed by significant but less firing throughout the rest of the stimulus; 3) sustained, in which the firing rate was essentially the same throughout the entire duration of the stimulus; and 4) primary-like with notch, a primary-like response but with a pause in firing just after an onset response and before a sustained response. Examples of these 4 response patterns are shown in Fig. 5.
|
CF and threshold
Characteristic frequency and threshold were not correlated (r = –0.080; Fig. 6A). Characteristic frequencies ranged between 0.9 and 6.1 kHz. Mean (±SE) CF for all cells was 3.46 (±0.14) kHz. Thresholds ranged between 4 and 75 dB SPL, with a mean of 34.82 (±0.17). Figure 6B shows the distribution of thresholds for cells grouped by response pattern. The distribution of thresholds was slightly wider for ongoing responses (4 to 75 dB SPL) than for onset responses (15 to 70 dB). The mean threshold for cells with onset responses (40.33 ± 2.09 dB SPL) was significantly higher (unpaired t-test; P = 0.005) than that for cells with ongoing responses (30.74 ± 2.12 dB SPL). There were no significant differences among the mean thresholds for cells showing primary-like, sustained, and primary-like with notch type responses (single-factor ANOVA; F = 1.714; P = 0.190). Mean (±SE) thresholds were 27.00 (±2.52) dB SPL for cells with primary-like responses, 35.92 (±5.31) for sustained responses, and 30.85 (±2.83) for primary-like with notch responses. Because many more cells with ongoing responses had lower thresholds (<25 dB SPL) than did cells showing onset responses (Fig. 6), and because all cells showing spontaneous firing had ongoing response patterns, it is possible that cells with ongoing responses are generally more excitable than cells with only onset responses.
|
Width of frequency tuning
Tuning curves were generated from FRA plots. Tuning curve widths were highly variable across the population of cells. The narrowest tuning curve we encountered showed a maximum width of 0.5 kHz at 50 dB above threshold and a minimum width of 0.1 kHz at just above threshold. The widest tuning curve spanned the 7.0-kHz range of test frequencies at maximum and 1.0 kHz at threshold. Examples of tuning curves are shown in Fig. 7.
|
; Theunissen and Doupe 1998). Tuning curve bandwidths were measured at 80 dB SPL and compared with temporal response patterns, thresholds, and CFs. There was no clear relationship between tuning curve bandwidth at 80 dB SPL and temporal response pattern or threshold. Characteristic frequencies were positively correlated with tuning curve bandwidth at 80 dB (r = 0.356, P = 0.0005); cells with wider tuning curves had higher CFs (Fig. 8). There was no difference in correlation between bandwidth at 80 dB SPL and CF in cells with onset responses versus cells with ongoing responses.
|
; Young and Brownell 1976) or type O (Ramachandran et al. 2000)] were observed.
|
|
To examine intensity coding, we collected RLFs in response to pure tones at CF. Cells that showed ongoing responses (n = 47) were the focus of this analysis because there was little or no change in spike rate as a function of intensity (above threshold) for cells showing only onset responses. Cells with ongoing responses fell into 3 classes in terms of their RLFs: 1) monotonic, in which spike rate continually increased as intensity increased (Fig. 11A); 2) low saturation, in which the spike rate increased with increasing intensity at lower intensities but then reached a plateau at which spike rate remained constant with further increases in intensity (Fig. 11B); and 3) nonmonotonic, in which spike rate increased with intensity up to some point, but then decreased with further increases in intensity (Fig. 11C). Of those cells showing ongoing temporal responses, 51% showed monotonic RLFs, indicating sensitivity to a wide range of changes in intensity, and 39% showed low saturation RLFs. Ten percent of cells (5/47) showed nonmonotonic RLFs. These cells had either primary-like or sustained temporal response patterns.
|
|
First-spike latencies were measured for responses to tones presented at CF, 20 dB above threshold. Under these conditions, the minimum response latency observed for the cells in our sample was 4 ms and the maximum latency was 40 ms. The mean first-spike latency for all cells was 11.36 ± 0.9 ms. Figure 13 shows the distribution of first-spike latencies for onset and ongoing responses, and the relationship between first-spike latency and CF. Most latencies were clustered around 10 ms (Fig. 13, A and B). No relationship between first-spike latency and CF was found (Fig. 13C).
|
For 15% of the cells in our sample there were no significant changes in spike latency across intensity. For the other 85%, response latency varied with both intensity and frequency. For 69% of all cells, latencies decreased with increasing sound level. Figure 14A shows a representative example of how response latency changed with sound level for an onset cell. In this example, the mean first-spike latency decreased from 8.1 to 6.0 ms as the sound level increased from 35 to 90 dB SPL. Figure 14B shows data from a cell with an ongoing response. For this cell, the mean first-spike latency decreased from 15.2 to 4.8 ms between 30 and 90 dB. For 16% of cells, latency decreased with increased sound level up to some point, but then increased with further increases in sound level. This effect is called a paradoxical latency shift because it is the opposite of the relationship usually observed between intensity and latency (Sullivan 1982). Figure 14C shows an example of paradoxical latency shift. Note that latency does decrease with increased sound level near threshold, but the pattern reverses at higher levels.
|
|
Inhibition
Because MLd cells in the anesthetized zebra finch show little or no spontaneous activity, we were not able to observe direct evidence of how inhibition shapes frequency and intensity coding in these neurons. However, the influence of inhibition on response characteristics was indirectly observed. First, the observation of complex tuning curves suggests the convergence of multiple excitatory inputs and or competing excitatory and inhibitory inputs onto a cell. Second, some cells showed nonmonotonic RLFs (Fig. 11), suggesting that higher sound levels recruited inhibitory inputs that suppressed the excitation observed at lower sound levels. Third, the observation of paradoxical latency shift suggests that inhibition can suppress the early part of the response at high intensities. Figure 16 shows a single cell's responses to tones presented at CF, at varying sound levels. The response exhibits both a nonmonotonic RLF and a paradoxical latency shift, suggesting the influence of inhibition on spike rate and response timing. At higher intensities, firing occurs after the offset of the stimulus. A cessation of firing accompanies the stimulus offset and is then followed by a resumption of firing, suggesting inhibition at the stimulus offset and the subsequent release from inhibition.
|
We tested cells at different sound durations to determine whether their responses were locked to the onset or offset of sound. Responses to sound offset are a characteristic of those cells in the mammalian IC that respond selectively to narrow ranges of sound duration. Pure tones of durations between 5 and 300 ms were presented at CF and at other frequencies. We found neither cells that responded to sound offset only nor cells that responded selectively to sound duration.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Single MLd units in the zebra finch showed numerous temporal response patterns and a wide range of frequency tuning, with some cells showing highly complex frequency coding. A general lack of sensitivity to intensity changes was observed. Many cells showed only onset responses that have small dynamic ranges and 39% of the cells with ongoing responses showed low-saturation RLFs. These findings suggest that the songbird auditory midbrain may be particularly good at processing temporally complex signals that show a high degree of AM. Here, we compare the tuning properties of zebra finch MLd units with those of auditory neurons in the midbrains of other birds, the inferior colliculus of mammals and the lower brain stem nuclei of birds. The implications of pure tone response characteristics for the processing of complex stimuli such as songs and calls are also discussed.
Comparison with MLd studies in other birds
Among birds in general, very little study has been devoted to auditory midbrain and thalamus (see Koppl et al. 2000). The one exception to this paucity of information is the auditory space map in the MLd (called the IC) of barn owls (Knudsen and Knudsen 1983; Knudsen and Konishi 1978; see Knudsen 1999 for review). Most of the work in owls has focused on mechanisms of sound localization, but some work has examined frequency and temporal tuning properties (Keller and Takahashi 2000; Knudsen and Konishi 1978; Koppl 1997, 2001; Sachs et al. 1980; Wagner 1990; Wagner et al. 1987, 2002). Neurons in the region of the barn owl midbrain that receives direct input from the cochlear nuclei, the ICCcore, share numerous tuning characteristics with the cells from which we recorded. Neurons in this region are organized into isofrequency layers (Takahashi and Konishi 1988) such that CFs increase along the dorsoventral axis (Knudsen and Konishi 1978; Wagner et al. 1987) and show single-peaked, V-shaped excitatory tuning curves (Wagner et al. 2002). Also consistent with our findings is the fact that the CFs of neurons in the ICCcore are distributed throughout the bird's entire hearing range (Wagner et al. 2002). This is in contrast to neurons in the ICx, a midbrain region just lateral to the ICCcore where cells respond only to higher frequencies and are tuned to specific interaural time differences (Knudsen and Konishi 1978; Mazer 1998; Takahashi and Konishi 1986). Response patterns in owl IC neurons can vary depending on factors such as changes in interaural time differences (ITDs; Wagner 1990). The sounds we used were presented free-field, from a speaker in front of the bird. Therefore we were not able to examine the effects of varying sound source location on temporal response patterns. Consistencies between the tuning properties of ICCcore units and our findings suggest that the cells from which we recorded in zebra finch MLd may correspond only to the ICCcore and not to the additional, more specialized regions in the barn owl auditory midbrain.
Two studies have examined tuning in MLd neurons of terrestrial birds. Coles and Aitkin (1979) examined response patterns, frequency tuning, and spike latencies in the chicken MLd. Similar to our findings, they found that onset and primary-like response types were common whereas offset responses were rare. As with the majority of our cells, latencies generally decrease with increasing intensity. Minimum latencies in the chicken were measured at 5 ms, agreeing well with our minimum latencies of 4 ms. Frequency tuning appears to be more complex in the zebra finch. Although the majority of MLd neurons in both species show V-shaped tuning curves, zebra finch neurons show complex tuning curves more frequently and with more complicated patterns. Coles and Aitkin found that 12% of chicken units showed double-peaked or "O-type" curves. We found complex tuning curves with multiple peaks, noncontiguous excitatory regions, and narrow, tilt, and columnar shapes in 22% of zebra finch units. Assuming that the coevolution of auditory and vocal systems affected their physiological functioning, differences in the complexity of frequency tuning could be related to differences in the acoustics of vocal signals. Chickens have relatively simple vocal repertoires that do not include learned components, whereas zebra finch vocalizations are complex in both frequency and amplitude, and are shaped by learning. Alternatively, it is possible that other factors that are unrelated to vocal behavior explain differences in frequency tuning. For example, chickens have smaller hearing frequency ranges than do song-birds. It is therefore possible that a decreased range of potential tuning frequencies results in simpler tuning curves.
Scheich and colleagues (1977) examined the responses of MLd neurons to vocalizations and some tonal stimuli in the guinea fowl. Their birds were unanesthetized and most of the neurons they investigated showed spontaneous activity. They divided cells into "simple" and "complex" in terms of how well a neuron's frequency tuning predicted its responses to vocalizations. Similar to many cells from which we recorded, guinea fowl units showed single- and multipeaked tuning curves, with tuning bandwidths generally widening with increased sound level. Cells that were nearly purely inhibitory, based on suppression of spontaneous activity, were also found. We would not have detected such inhibitory neurons in our sample because our cells had no spontaneous rate. Characteristic frequencies fell between 1 and 4 kHz, indicating correlations between CF range, hearing frequency range, and the power spectra of guinea fowl vocalizations. Similarly, the range of CFs in our cells (0.9–6.1 kHz) match both the zebra finch audiogram (Okanoya and Dooling 1987) and the power spectrum of zebra finch song (S.M.N. Woolley, personal observation). The fact that Scheich et al. found "complex" units for which tonal responses could not predict responses to natural calls provides evidence in favor of the idea that the MLd may contain nonlinear tuning properties that correspond to a bird's vocal patterns.
Comparison with mammalian auditory midbrain— is the bird different?
The mammalian auditory midbrain region, the inferior colliculus (IC), has been extensively studied in a variety of species including cats, chinchilla, and mice. Here, we compare what is known about response patterns, frequency tuning, and interactions between latency and sound level in the IC to our findings in the MLd. Clear similarities among the response patterns found in the mammalian IC and MLd exist. As in MLd, some mammalian IC cells exhibit only onset responses to tones; the percentages of total cells showing only onset responses varies across species and studies (Egorova et al. 2001; Nuding et al. 1999; Pollak et al. 1978; Rose et al. 1963; Skya et al. 2000). However, the IC appears to be less dominated by onset cells than the MLd (although see Wagner 1990). Tonic responses of the primary-like, sustained, and primary-like with notch types (as defined here) are also included in the mammalian IC (Pollak et al. 1978; Skya et al. 2000). Some studies have shown that additional response patterns such as onset–offset, offset, and purely inhibitory are present in the IC (Casseday et al. 1994; Ehrlich et al. 1997; Pollak et al. 1978; Skya et al. 2000). We did not find onset–offset or offset units. Mammalian onset units respond with latencies that are shorter and more temporally constrained than those of cells with ongoing responses (Nuding et al. 1999). The fact that onset cells appear to be more temporally consistent in both birds and mammals suggests that onset cells in general may be best at encoding some aspects of time-varying signals because the timing of changes in both frequency and amplitude can be accurately preserved in the neural response.
Our findings suggest that the zebra finch MLd shows the same general pattern of tonotopy as is found in the mammalian IC (Casseday and Covey 1992; Nuding et al. 1999; Semple and Aitkin 1979). Consistent with observations in the MLd of the guinea fowl (Sheich et al. 1977) and the IC of cats and bats (Casseday and Covey 1992; Pollak and Bodenhamer 1981; Pollak et al. 1978), CFs in the zebra finch correspond well with the frequency range and power spectra of species-typical vocal signals and species-specific audiograms. Many units in both cat and bat IC show V-shaped tuning curves similar to those that we observed in the zebra finch. However, the bat IC contains neurons that show more specialized frequency tuning characteristics than any we observed in the zebra finch. Some bat IC neurons show "narrow-filter" tuning curves, with extremely limited excitatory frequency response areas. Narrow filter neurons are found in the brain region that encodes frequencies contained in the narrowband echolocation signals that the bat uses when searching for prey (Casseday and Covey 1992; Pollak and Bodenhamer 1981). Some MLd units did show level-tolerant "columnar" frequency tuning curves that were similar to the narrow-filter units of bats but broader. Such tuning differences may reflect differences in the acoustics of the vocal communication signals in bats and zebra finches; bats use narrow-frequency signals and zebra finches use broadband signals.
Comparison with tuning properties of cochlear nucleus neurons—what properties may emerge in MLd?
The ascending inputs to MLd are the brain stem nuclei, nucleus magnocellularis (NM), nucleus laminaris (NL), and nucleus angularis (NA). These areas have been well described both anatomically and physiologically in chicks and owls (Carr and Konishi 1988; Koppl 1997, 2001; Koppl and Carr 2003; Parks and Rubel 1978; Rubel and Parks 1975; Sachs and Sinnott 1978; Sullivan 1985; Takahashi and Konishi 1988; Warchol and Dallos 1990). One study examined the frequency tuning found in NM and NA of a songbird, the red-winged blackbird (Sachs and Sinnott 1978). Some basic response characteristics are shared between MLd and the cochlear nuclei of other birds. A match between hearing range and the range of CFs measured in the MLd is also found in the cochlear nuclei (most clearly in NA) of chicks, red-winged blackbirds, and barn owls (Koppl and Carr 2003; Rubel and Parks 1975; Sachs and Sinnott 1978; Warchol and Dallos 1990). The wide range of thresholds found in MLd is consistent with the range of thresholds found in NM and NA (Koppl and Carr 2003; Warchol and Dallos 1990). The average threshold sensitivity of MLd neurons appears to match better with NA than NM; NM thresholds are generally higher. In the cochlear nuclei and in MLd, no correlation between CF and threshold is found.
In all of these bird species, NM contains only one, primary-like, response pattern. NA contains primary-like and onset responses (Koppl and Carr 2003; Sachs and Sinnott 1979; Sullivan 1985; Warchol and Dallos 1990) as well as choppertransient responses (Koppl and Carr 2003; Sullivan 1985; Warchol and Dallos 1990). We recorded some chopper-transient responses in MLd of the zebra finch (see Figs. 5E and 14B), but the periodicity we observed was not generally as pronounced as that reported for NA. Although we performed many tests on an individual cell, each test typically contained only 15 to 25 trials. This number of repetitions provided enough data to classify units into major categories (given that multiple tests were used), but the data were not sufficient to distinguish chopper-like from primary-like response patterns. Therefore these responses, which showed the overall shape of primary-like responses, were classified as primary-like. The sustained responses found in 20% of MLd units and the primary-like with notch response types found in 12% of MLd units have not been described in avian cochlear nuclei, although they are present in the mammalian auditory midbrain, nuclei of the lateral lemniscus, and cochlear nuclei (Covey and Casseday 1991; Pollak et al. 1978; Skya et al. 2000). This suggests that, in birds, the sustained and primary-like with notch response types that are typical throughout the mammalian auditory system are first seen in the MLd.
Another notable difference between MLd and the cochlear nuclei in terms of temporal response patterns is the preponderance of temporally precise, onset responses in MLd. Although onset responses are present in NA, they do not occur in high numbers (Koppl and Carr 2003; Sullivan 1985; Warchol and Dallos 1990). Additionally, NA cells do not give a purely onset response; responses are usually strong to the onset of a stimulus and weakly continued to the rest of the stimulus. Nearly all onset cells in MLd respond only to the stimulus onset and do not spike again until a second stimulus is presented. In this way, MLd onset units may be particularly good at marking the temporal relationships between sequential stimuli with high fidelity, potentially firing and recovering quickly to encode rapid amplitude modulations.
Tuning curves for MLd cells appear to differ significantly from those found in the cochlear nuclei. Tuning curves in cochlear nucleus neurons are V-shaped, apparently without a wide variety of bandwidths. We observed V-shaped frequency tuning curves in 78% of units, but the Q10s, Q30s, and bandwidths at 80 dB SPL varied greatly. Our V-shaped curves showed maximum frequency bandwidths as limited as 1 kHz and as broad as 7 kHz. Additionally, 22% of the MLd tuning curves were complex, showing multiple peaks and/or multiple excitatory regions, columnar shapes, and tilt shapes. The fact that these types of complex frequency tuning curves have not been found at lower levels suggests that they emerge in MLd. The presence of GABAergic neurons has been documented in the avian MLd (Carr et al. 1989; Fujita and Konishi 1991; Granda and Crosland 1989), suggesting that GABAergic inhibition might play a role in generating complex frequency tuning through the interplay of excitatory and inhibitory inputs. In mammals, especially in bats, multiple peaked tuning curves are found in IC neurons that respond optimally to 2 tones that differ in specific frequency and temporal spacing from one another (Dear and Suga 1995; Feng et al. 1978; Portfors and Wenstrup 1999, 2001; Simmons 1971, 1973). This raises the question of whether similar frequency-combination, delay-sensitive neurons exist in MLd.
MLd units were found to have limited dynamic ranges, meaning that they were relatively insensitive to changes in sound level. Onset cells (49% of MLd cells) inherently show small dynamic ranges because of the fact that very few spikes are elicited by any one stimulus. Many of the MLd cells with ongoing responses showed RLFs that saturated at relatively low intensities, with spike rates plateauing between 40 and 50 dB SPL. In total, only 26% of all the cells from which we recorded showed rate-level functions that did not saturate at our highest test intensities. The level insensitivity to sounds presented at intensities over approximately 50 dB SPL that was exhibited by most MLd neurons is in contrast to the RLFs described for neurons in the cochlear nuclei. Few cells in NM and NA are reported to show low-saturation RLFs. The functional consequences of an emergent insensitivity to intensity changes in MLd remain to be examined. Such cells may function best while coding signals with significant AM. It is possible that consistent responses regardless of overall intensity differences could act to stabilize the precise encoding of temporally complex signals that have the same meaning regardless of intensity. It is important to note, however, that the frequency bandwidth to which an MLd cell responds is highly dependent on intensity. Thus complex interactions between the intensity and frequency modulations that are typical of biologically meaningful sounds would still be expected to occur in MLd.
Implications for the processing of natural, complex sounds
The limitations of using simple, synthetic stimuli such as pure tones to examine how auditory circuits process sound in general have been recognized for decades (Casseday and Covey 1996; Pollak et al. 1978; Scheich et al. 1977; Theunissen and Doupe 2000). Nonlinear tuning properties that represent sensory adaptations (specializations) for coding vocal signals may be revealed only by approaching the system using complex stimuli, preferably containing the acoustic properties of the natural sounds under which the auditory system has evolved. In some systems, a unit's responses to 2 tones presented separately cannot predict the response to those same tones presented simultaneously or sequentially (Dear and Suga 1995; Jen et al. 2002; Nieder and Klump 1999), much less within a complex signal (Klug et al. 2002; Scheich et al. 1977). Thus studying the responses of MLd neurons to pure tones may fail to reveal tuning complexities in the songbird midbrain. With this caveat in mind, it is helpful to have a basic understanding of the responses of these neurons to simple stimuli in terms of response patterns, and frequency/intensity coding before attempting to understand their responses to complex stimuli. Here, we draw 3 conclusions about the processing of simple stimuli that allow us to make predictions about how these neurons may function in processing complex signals such as songs and calls. In this way, we can ask what information within natural sounds may be transmitted to higher levels through MLd.
Half of the cells we encountered showed only onset responses to free-field sound presentation, with precise timing regardless of intensity changes. These cells could be optimal for encoding the temporal patterns of complex signals that are characterized by amplitude modulations such as songs. For example, a broadband onset cell could detect the onsets of any vocal element that is preceded by silence, such as a song syllable, thereby preserving the amplitude envelope of the song through a distinctly patterned sequence of spikes. Narrowband onset cells could detect the onsets of particular vocal elements that are not necessarily preceded by silence but that contain frequencies falling within the excitatory tuning curve of that cell. Second, we found a large variety of frequency tuning characteristics across cells. A population of cells showing highly different frequency tuning could be beneficial for generating activity in a unique constellation of neurons in response to each call, song note, song syllable, and so forth. In this scenario, the concurrent firing of a specific group of midbrain neurons could serve as the neural representation of an individual vocalization. Finally, most of the neurons we characterized were surprisingly insensitive to i
1.5 kHz width at 80 dB SPL) tuning curves spanned a range of thresholds and CFs. B: midwidth (between 1.5 and 4.0 kHz width at 80 dB SPL) tuning curves, also spanning a variety of thresholds and CFs. C: wide (
