Verification of recording sites
Six recording sites were marked by electrolytic lesions. Five of these were later identified in histological sections of NA (examples in ). The sixth lesion could not be found, presumably because the animal did not, as intended, recover from the experiment, and the survival time was too short to reveal the characteristic glial accumulation. One further track was approximately confirmed by an ink mark placed on the surface of the brain stem. The tonotopic position of lesion sites corresponded with the best frequencies estimated immediately before current application. The tracks where the five successful lesions were placed accounted for 12 of the units reported here; all of those were classified as NA and all response types [except the rare chopper-sustained (chop-S)] were represented. Using the lesioned sites as calibration points and the stereotaxic coordinates of electrode penetrations in the same individual as guidelines, we could reconstruct the approximate rostrocaudal position for about 70% of all recorded units. These recordings covered the rostral two-thirds of NA. The remaining units either came from brain stem hemispheres without any anatomical confirmation of recording sites or from penetrations aimed using landmarks. In the mediolateral axis, a detailed reconstruction of anatomical position was not possible. In this axis, the nucleus is much narrower and electrode angulation introduced an additional variable that could not be measured and reproduced as precisely as the electrode's X and Y positions. However, characteristic frequency changes along the mediolateral axis of NA in the barn owl (Köppl 2001
), such that the lowest CFs can only be found at the most medial positions and the highest CFs at the most lateral extreme. Judging by this criterion, we have sampled the entire medio-lateral extent of NA (see data in the following text).
FIG. 1 Two examples of lesions that were placed to mark recording sites and were subsequently found in nucleus angularis (NA). A and C: drawings of the outlines of the brain stem and several auditory nuclear areas in a transverse section. The areas of lesions (more ...)
In summary, all recording sites that could be confirmed anatomically were confirmed to have been within NA, corroborating our judgements based on physiological criteria during the experiment.
Classification of response types
Responses of a total of 76 single units recorded in the NA are reported here, covering a range of CFs from 0.45 to 10.2 kHz. The responses clearly were not uniform but could be classified into several different categories. Because they appeared to be comparable to types previously defined in the mammalian cochlear nucleus, we have adhered to the established nomenclature. summarizes the types, their relative abundance, and several salient statistics of their response behavior. presents the decision tree that was developed, using a range of criteria, including the traditional PSTH parameters. The most common response of NA neurons was of the primary-like variety, followed by chopper-transient (chop-T), typeIV, onset, and, rarest, chop-S responses.
Summary of salient characteristics of the different response types
Decision tree developed for classifying the recorded units.
Note that in addition to the sample of NA units, there are 26 units that were classified as “auditory nerve” or “uncertain” (). This reflects some difficulty in separating potential auditory-nerve responses from primary-like NA units. We assumed that our sharp tungsten electrodes were capable of recording from both cell bodies and large axons or dendrites (see e.g., Joris 1998
) and that all of these could in principle be encountered anywhere within the nucleus. Therefore a conservative interpretation was used, classifying all units whose values for a list of parameters fell within the known range of auditory-nerve fibers (Köppl 1997a
; Köppl and Yates 1999
) as “auditory nerve.” Based on previous studies on the avian NA (Sachs and Sinnott 1978
; Sullivan and Konishi 1984
; Warchol and Dallos 1990
) and our own auditory-nerve data from the barn owl (Köppl 1997a
), a relatively low vector strength of phase locking and/or a relatively low spontaneous discharge rate were considered decisive parameters for separating NA units from auditory-nerve inputs. Note that no specific criterion values can be universally used for this distinction because both parameters are strongly CF dependent. For example, a unit with a CF of 5 kHz and a vector strength of 0.5 would have been classified as auditory nerve, whereas a unit showing the same vector strength at a CF of 2 kHz would have been classified as NA. Cases where both the vector strength and the spontaneous discharge rate were consistent with auditory-nerve responses, but other characteristics did not fit an auditory-nerve fiber (e.g., a high maximal discharge rate) were classified as “uncertain.” Latency was, unfortunately, of little use in unit classification (see following text, after introducing the response types).
We will first describe and illustrate with examples the typical features of each NA response type. Then the different types will be compared and salient differences and similarities highlighted. It should be emphasized that most of the response types did not appear to be sharply distinguished, i.e., their characteristics overlapped to some degree for individual parameters. In addition, many parameters showed an overall CF-related variation that we have tried to separate and disregard in our classification of response types. A number of scatter plots will be shown in addition to individual examples to illustrate the range of responses and the extent of overlap between types.
Response type primary-like
illustrates an example of a primary-like response. The characteristic feature of primary-like units was, of course, a primary-like PSTH, showing a vigorous discharge at stimulus onset that gradually adapted to a steady, lower discharge (). Occasionally, the PSTH showed a clear notch after a peak at response onset. However, this was confined to high levels of 40 dB or more above threshold and intermediate forms were seen. Also, such notches may sometimes be observed in auditory-nerve responses at high levels (own unpublished observations). Therefore we do not feel there is evidence for a distinct response type “primary-like with notch” (however, see also “onset units” in the following text). Interspike interval distributions of both spontaneous and evoked discharges were Poisson-like (). Mean CVs were around 0.8 – 0.9 (median: 0.84), except in low-frequency units where cycle-by-cycle phase-locking produced a more regular response.
FIG. 3 Example for a unit with primary-like response. A: tuning curve. B: rate-level functions for stimulation at characteristic frequency (CF) = 10 kHz and with noise. C: peristimulus time histograms (PSTH) at CF at 80 dB SPL = 38 dB above threshold, and ( (more ...)
All primary-like units with a CF less than 5 kHz showed significant phase locking. Although their vector strengths were generally lower than those of auditory-nerve fibers of comparable CF, they could approach auditory-nerve values at CFs less than 1.5 kHz (). At CFs above 5 kHz, vector strengths were very low or not statistically significant. Spontaneous discharge rates were mostly below those of auditory-nerve fibers of comparable CF ( and ).
FIG. 4 Vector strength for stimulation at CF, at least 20 dB above threshold, as a function of CF. Data for the different response types are drawn with different symbols as indicated; the dot data points and dashed line serve as a reference, displaying published (more ...)
FIG. 5 A: spontaneous discharge rate as a function of CF. Data for the different response types are drawn with different symbols as indicated; the dot data points and dashed line serve as a reference, displaying published data and an exponential fit for auditory-nerve (more ...)
Tuning curves of primary-like units were comparable to auditory-nerve tuning curves, showing a similar range of thresholds, Q10 dB and Q40 dB values. Four of 33 primary-like units showed evidence for side-band inhibition along the high-frequency flank of the excitatory tuning curve.
Rate-level functions of primary-like units were monotonic (), with a median saturation discharge rate of 257.5 spikes/s and a median dynamic range of 30 dB (). Responses to noise were similar to pure-tone responses, with respect to PSTH (), regularity of discharge and saturation rates.
Response type Chop-T
and illustrate two examples of chop-T responses. Chop-T units showed a regular discharge pattern at stimulus onset that produced several distinct peaks in the PSTH ( and ). This did not depend on the stimulus rise time in three cases tested (rise time varied between 1 and 5 ms; example in ). Interspike intervals always increased within the first 10–20 ms of the stimulus ( and ); the CV mostly decreased but could also remain steady or increase slightly ( and ). Mean CVs fell between 0.38 and 0.84 (median: 0.53). At the upper end of this range, there was some overlap with primary-like units. Besides the characteristic peaks in the PSTH, chop-T units also typically did not show a Poisson-like distribution of interspike intervals in their sound-evoked discharge. Their distributions were more skewed, with an increased proportion of short intervals ( and ).
FIG. 6 Example of a unit with a chopper-transient (chop-T) response, showing a high saturation discharge rate. The layout is the same as in . A: tuning curve. B: rate-level functions for stimulation at CF = 6 kHz and with noise. C: PSTH at CF at 70 dB (more ...)
FIG. 7 Example of a unit with a chop-T response, showing an average saturation discharge rate. The layout is similar to . A: tuning curve. B: rate-level functions for stimulation at CF = 7.4 kHz and with noise. C: PSTH at CF at 40 dB SPL = 34 dB above (more ...)
Spontaneous rates of chop-T units were typically low, with many units showing no spontaneous discharge at all ( and ). Chop-T units were also characterized by inferior phase-locking. Although about half of them did show significant phase-locking, their vector strengths remained at the low end of those of primary-like units of comparable CF ().
Chop-T units had a median saturation discharge rate of 316 spikes/s and a median dynamic range of 23 dB but showed large ranges for both parameters (). There was no correlation between saturation rate and dynamic range. Discharge rates and dynamic ranges in response to CF tones and to noise, respectively, showed no systematic differences across the population. However, the PSTH of noise responses never showed clear chopping ().
Chop-T units at the low end of the saturation discharge range showed PSTH with little sustained activity after the initial chopping peaks. Therefore we earlier contemplated a separate response type, “onset-chopper.” However, a more detailed analysis did not produce any evidence for this in the owl. Onset-choppers in cats and guinea pigs are characterized by their large dynamic range, small difference in threshold to tones and noise, and often more vigorous response to noise (Joris and Smith 1998
; Rhode and Smith 1986
; Winter and Palmer 1995
). While these individual characteristics were occasionally observed in our chop-T units (e.g., small threshold difference in ), none displayed all of them. Also, the PSTH in response to noise never showed chopping, while the PSTH of onset-choppers are very similar in response to tones and noise (Winter and Palmer 1995
An unexpected phenomenon was a decrease in spike size during medium- to high-level stimulation, observed in about one quarter of all our chopper-type neurons (including chop-S, an example of which is shown in ). In early experiments, we may have rejected some units like these, mistaking the nonuniform spike size as indication for a multi-unit recording. However, the presence of a refractory period in the interspike-interval histogram confirmed that the spikes were from single units despite the variable spike height.
FIG. 8 Example of one of the rare chopper-sustained (chop-S) units. The layout is similar to . A: tuning curve. B: rate-level functions for stimulation at CF = 6.3 kHz and with noise. C: PSTH at CF at 50 dB SPL = 26 dB above threshold and (inset) 40 dB (more ...)
Response type chop-S
Chop-S responses were rare in our sample and are only represented by two units (from different animals). One example is shown in . They had the longest response latencies and were clearly above all other unit types in this respect (). This was our main criterion for placing them in a category of their own, instead of interpreting them as one extreme of the chopper-type range.
FIG. 9 Minimal response latency to stimuli at CF at 20–35 dB above threshold, as a function of CF. Data for the different response types are drawn with different symbols as indicated. Each unit is only represented once. Note the extensive overlap between (more ...)
Chop-S units showed the lowest mean CVs of all types and their CV remained constant throughout the stimulus duration (). Also, in one case the unit chopped to both pure-tone and noise stimuli (; in the other case, the noise spectrum level was not sufficiently above threshold). Both chop-S units had no spontaneous activity and reached only moderate saturation discharge rates around 200 spikes/s. Their dynamic ranges were rather different, at 16 and 35 dB, respectively.
Response type onset
Units that showed only a single prominent, initial peak in their PSTH and no evidence for chopping, were classified as onset. A second characteristic was that the maximal discharge rate to noise was higher than that to tones at CF. In other aspects, this group was heterogeneous; however, considering the overall rarity of onset units and thus small sample size (n
= 6), we refrain from further subdividing. and illustrate the range of responses. The unit shown in represents an extreme case of onset response with virtually no spiking during the remainder of the stimulus. The unit illustrated in showed a robust sustained response after the onset and was in some aspects reminiscent of primary-like-with-notch responses in mammals. However, the distinction between onset and primary-like-with-notch is not always sharp in mammals as well (e.g., Blackburn and Sachs 1989
; Rhode and Smith 1986
). Our final criterion for classifying this (and another similar unit) among onset responses was the clearly more vigorous response to noise. The SD of the first spike, a measure for the synchrony of the initial discharge, was extremely low for the unit shown in but, as a population, actually larger in onset units than in all other types. However, this mainly reflects the fact that the onset spike occasionally failed, producing a distribution of first-spike times with extreme outliers and thus making SD an inappropriate measure of dispersion. If an onset spike was fired, it was temporally precise, summing up, over many stimulus repetitions, to the characteristic sharp peak in the PSTH. The PSTH in response to noise was more primary-like in all cases tested, i.e., showed less synchronization to the stimulus onset ( and ). Onset units did not appear to phase-lock well (), however, our sample is small and restricted in CF.
FIG. 10 Example of a unit with an onset response, showing little sustained discharge. A: tuning curve. B: rate-level functions for stimulation at CF = 4.7 kHz and with noise. C: PSTH at CF at 75 dB SPL = 42 dB above threshold and (inset) 55 dB SPL = 22 dB above (more ...)
FIG. 11 Example of a unit with an onset response, showing a robust sustained discharge as well. A: tuning curve. B: rate-level functions for stimulation at CF = 4.0 kHz and with noise. C: PSTH at CF at 40 dB SPL = 25 dB above threshold and (inset) 20 dB SPL = (more ...)
Eleven units were classified as typeIV, with CFs between 0.6 and 7.7 kHz. Their defining characteristic was a pronounced nonmonotonic behavior of the discharge rate across different levels at CF, typically showing little to moderate excitation at low levels and inhibition at higher levels (). The PSTH showed a clear onset response, followed by varying degrees of inhibition, depending on the unit and the stimulus level; we call this pattern onset-inhibitory. Units that responded with net excitation at low levels showed a sustained excitatory response at those levels, i.e., the PSTH changed with stimulus level (, and ). At levels near the transition between net excitation and net inhibition, the PSTH could also look pauser-like. When testing a range of frequencies and levels, complex response maps resulted, with interleaving excitatory and inhibitory areas (). CF and threshold were defined in these cases as the most sensitive point, regardless of whether the response was net excitatory or inhibitory. All typeIV units had high spontaneous discharge rates, typically above 100 spikes/s ( and ).
FIG. 12 Example of a unit with a typeIV response. A: response map. The areas enclosed by solid black lines delineate excitatory stimuli, the gray line indicates the lower boundary of the inhibitory area. The 3 dots indicate the frequency and levels of the stimuli (more ...)
FIG. 13 Example of a unit with an extreme typeIV response, showing no inhibition below spontaneous rate. A: response map. There was a V-shaped band of excitation (—). ●, the frequency and levels of the stimuli used for the PSTHs in C–E. (more ...)
Most typeIV units were also tested with noise stimuli. The majority showed purely excitatory responses and a monotonic increase in discharge rate with level (). This was accompanied by a primary-like pattern in the PSTH (). Three units gave similar responses to noise and tones, i.e., a net inhibition at higher levels, with an onset-inhibitory PSTH ().
Inhibition was seen to varying degrees in the different type IV units. At one extreme were units that showed little or no net inhibition, i.e., whose discharge rate never clearly decreased below the spontaneous rate (example in ). At the other extreme were units that showed no clear excitation, i.e., whose only response appeared to be a net decrease of the discharge below spontaneous rate. However, the onset response in the PSTH was always present (example in ). This was our final criterion for not placing such units into a (purely inhibitory) category of their own. It was also our impression that the responses recorded from typeIV units had a snapshot character, meaning that the degree of inhibition seen in an individual neuron could change over time. Significant changes in the discharge behavior over time could be documented for three single units. One example where inhibition became more pronounced with time is shown in . Stability or instability of the inhibitory response component did not appear to be related to the administration of anesthetic agents.
FIG. 14 Example of a unit with an extreme typeIV response, showing no excitation above spontaneous rate. A: inhibitory tuning curve, drawn in gray. The 3 dots indicate the frequency and levels of the stimuli used for the PSTHs in C–E. B: rate-level functions (more ...)
FIG. 15 Example of a change in the rate-level function of a typeIV unit. Two functions are shown that were recorded at CF (700 Hz) under identical conditions 35 min apart. Note that the spontaneous rate (- - -) did not change. However, threshold shifted slightly (more ...)
Comparison of types
Our sample covered the full range of CFs expected for the barn owl, although only primary-like units were found throughout the whole range. Especially conspicuous was the absence of any type other than primary-like among the highest CFs, more than 8 kHz. Chop-T units, although the second most frequent type, were also restricted at the low-frequency end and were only encountered at CFs between about 2 and 8 kHz. Anatomically, we did not observe any differences in where the various response types were found.
Perhaps surprisingly, latency was not clearly different between units classified as auditory nerve or NA, respectively, or between most of the NA response types. A number of variables influence latency, most importantly the CF and, using a fixed rise time, both the absolute and relative level (above threshold) of the stimulus (e.g., Heil and Neubauer 2001
). It seemed that the scatter introduced by those variables largely obscured the small latency differences between auditory-nerve fibers and NA units (). The only exception with clearly different latencies were the chop-S units. Higher sound levels or clicks might have provided more consistent information, but unfortunately unit isolation could be compromised with such stimuli. We also evaluated mean and median first-spike latency, and these showed greater variability than the minimal latency shown in .
Except for the typeIV, the different response types appeared to grade into each other. The extremes of the primary-like and chop-T responses, for example, were clearly different, however, a few units could only be classified by defining an arbitrary borderline value for the mean CV and subjectively classifying their PSTH shape. Similarly, chop-T units with low saturation discharge rates may form a continuum with onset units.
A number of parameters did not differ at all between most types of units. Tuning curves largely reflected the auditory-nerve inputs in terms of tuning and thresholds (). Only a minority of primary-like and chop-T units (4 of 33 and 2 of 17, respectively) showed evidence for off-CF inhibitory inputs along the high-frequency flank of their excitatory tuning curves. TypeIV units tended to fall among the most sensitive thresholds; however, there was no statistically-significant difference in threshold between the response types (Kruskal-Wallis H-test, P = 0.06; only units with a CF less than 8 kHz tested to minimize the variation of threshold with CF). Dynamic ranges were not significantly different between auditory-nerve units and the different NA response types (Kruskal-Wallis H-test, P = 0.22; typeIV units excluded because no dynamic range was defined for nonmonotonic rate-level functions). For the driven discharge range, i.e., the difference between spontaneous and saturation rate, significant differences were revealed. However, it was only the onset group that differed from all others (Kruskal-Wallis H-test with subsequent pairwise Mann-Whitney U-tests; typeIV units excluded). This is probably entirely due to the low saturation discharge rates of onset units, which similarly differed from those of all other types (see also ).
FIG. 16 A: threshold at CF as a function of CF. Data for the different response types are drawn with different symbols as indicated in B. - - -, a reference displaying a 2nd-order polynomial fit to the values of 335 auditory-nerve fibers (own partly unpublished (more ...)