Condor 85:259-26 I Q The Cooper Ornithological Souety 1983 (carrier frequency).The side bands will be separatedfrom COMMENTARY the carrier frequency and, if there are several, from each other by the frequency of repetition rate, or modulating frequency. Hence, if the carrier frequency can be evenly ON SONOGRAMS, HARMONICS, divided by the modulatingfrequency, the sonographicpat- AND ASSUMPTIONS tern will resemble a harmonic spectrum of a tone with a fundamental frequency equivalent to the modulating fre- The ability of sonogramsto present acoustic data in a quency. The carrier frequency usually appears as an em- format that is easily comprehended,compared, and print- phasized harmonic. Becausethe modulating frequency is ed has made them an important tool in studies of avian usually much lower than the carrier frequency, the ap- behavior and systematics.The 10 volumes of The Condor parent “fundamental frequency” is low, the “harmonic from 1972-198 1 contained 51 papers, involving 75% of spectrum” is composed of closely spaced bands, and, if all issues,that used sonographicdata and two papers dis- the carrier is high, the “lower harmonics” are usually ab- cussingthe presentationand use of such data. This com- sent. The exact sonographicpattern depends both on the ment follows in the latter series by indicating some pos- nature of the modulation and the setting of the spectro- sible pitfalls in the interpretation of sonograms. graph; the problem is acute with narrow band-pass filter Birds can utter a wide variety of phonations, but these settings.Appropriate soundsfor producing such patterns can be divided into two categoriesdepending on how the appear to be common in avian phonations (Davis 1964, soundis produced.The most common kind of avian pho- Watkins 1967, Marler 1969). nation is a simple tone, or whistle, in which all the energy Third, a freely oscillating,edge-clamped membrane, such at any given instant is concentratedin a single frequency. as a syringealmembrane, behaves quite differently from Sonogramsof unmodulated, simple tones show a single a freely oscillating,end-clamped string.A wave propagates horizontal band. The production of a whistled sound does in one dimension along the length of the string. If, as is not depend on the oscillation of a membrane or any other probable, the wave encountersan impedance mismatch mechanical element of the syrinx (Gaunt et al. 1982). A at the end of the string, some of it will reflect back in the second kind of sound is that in which, at any moment, opposite phase, so that it seems to have changed to the energy is distributed into more than one frequency. The opposite side of the string. If the vibration is repeated, generationof many of these soundsis supposedto involve succeedingwaves will interact with the reflected waves to either mechanical oscillatorsor coupled resonators.I am form a series of standing waves, the number of which is concernedhere with multi-frequency soundsin which the equivalent to the number of harmonics. In a membrane, energyis distributedinto distinct, discreetfrequencies, i.e., however, the wave radiates as a series of arcs from the not broad-band noises.Davis (1964) included suchsounds point of stimulation. Hence, it propagatesand is reflected among those he called “complex tones.” However, I find in two dimensions,deforming a surfacerather than a line. it useful to restrict “complex tones” to those soundsthat The resultinginteraction is far more complex than in the are produced by a single generator. A sonogram of an one-dimensional string, and the resulting overtones are unmodulated, complex tone contains a series of parallel not normally harmonic. Rather, they will occurat varying bands. Such sounds are popularly termed “harmonic fractional (partial) multiples of the fundamental (Rossing tones,” and the distribution of frequenciesis called a “har- 1982). Such partial overtones account for the character- monic spectrum.” In a true harmonic tone, such as is istic, pitchlesssounds of many drums. Tunable drums with producedby most stringedand wind instruments,the higher distinct pitch, suchas kettledrums,represent special classes frequencies(overtones) are integer multiples of the fun- of membrane instruments in which various factors force damental frequency (or first harmonic). This simple re- the partial frequencies closer to harmonic values of an lationship is familiar to many researchers,but some in- apparentlymissing fundamental. Casey (198 1) showedthat vestigators assume that all multi-frequency sounds are partials will be generated by membranes of a variety of harmonic tones. Therein lies a set of problems because(1) shapessubjected to various degreesof damping. Hence. it not all multi-banded figureson sonogramsrepresent com- is unlikely that a freely oscillating syringealmembrane of plex tones, (2) not all complex tones are composed of any shape, even if constrained like the head of a kettle- harmonics as here defined, and (3) the mechanism(s) drum, will producea classicharmonic spectrum.Yet some wherebya syrinx generatescomplex tonesis (are) not clear- birds, e.g., Phainopepla (Phainopepla nitens; Leger and ly understood. Carroll 1981) and Whimbrel (Numeniusphaeopus; Skeel First, an apparent harmonic may be an artifact. Sound 1978) do seem to utter soundswith true harmonic spectra. spectrographswill produce spuriousharmonics if the sig- This suggeststhat something is faulty, or at least lacking, nal is introduced at too high a gain (Davis 1964, Gree- in our present understandingof syringealmechanics. newalt 1968:9). This type of error is usually overcome Problems in the interpretation of complex tones may with experience, but that experience should not be as- be exacerbatedby improper use of a sonogram. For in- sumed in reviewers, editors, or readers. stance,the analysisof an harmonic spectrum depends on Multi-frequency soundsmay also be the product of the a rather precise determination of frequencies,but many two-voice phenomenon. If both sidesof the syrinx should sound spectrographs,including the commonly used Kay produce complex tones with slightly different fundamen- Elemetrics“Sona-Graph,” do not measurefrequency well, tals, very puzzling patterns may result (S. Nowicki, pers. especiallywhen used with a wide-band filter. The trigger comm.). The two-voice phenomenon is surelyresponsible frequency is generally assumedto lie at the center of the for instances in which frequency bands cross over each inscribed band, but that may not be so (Davis 1964) and other, diverge with falling frequencies,or converge with even if it does, the center must be estimated with some rising frequencies. (Overtones, whether partial or har- degreeof error. Further, the width of the band varies with monic, must diverge with rising frequenciesand converge the gain setting of the spectrograph.Even when using a with falling ones.) In a confusion of the conceptsof two- narrow-band pass filter, e.g., 40 Hz, for high resolution, voices and harmonics, Anderson (1978) described “pe- one should expect a production error of up to ?20 Hz. culiar harmonic bands” exhibiting several of the cri- To this must be addedthe investigator’sinterpolation error. teria for two voices, that may be an example of “dual Calibration marks are, at best, at 500 Hz, more usually control of sound production.” 1,000 Hz intervals, and intervening values must be inter- Second, if a tone is sufficientlymodulated in either fre- polated. On a sonogram with a frequency range of 80- quency or amplitude, then its sonogram will show side 8,000 Hz, which is common in avian studies, each mil- bands to both sidesof and parallel to the tone’s frequency limeter on the ordinate represents70 Hz. Hence, a mistake 12591 260 COMMENTARY of 10 Hz will result from an interpolation error of only Hand’s data probably did conform to that pattern. Her 0.14 mm! Figure 8 suggeststhat the Long Call of some gulls is com- For most behavioral or systematicinvestigations, failure posed of harmonics. The third, and in my view most se- to recognizethese problems is oflittle import, but for those rious flaw, is failure to state explicitly that at least some of us interested in syringeal function, such errors can be of her data are not easily fitted to the expected pattern. serious.The reason for this is that we have yet to devise On such discrepanciesscientific theories can hang or fall, a technique to observe a functioning syrinx in vivo. True, and failure to recognizeor proclaim them only preserves syringealstructures are well-known, but most of our no- the present paradigm. tions of how those structureswork are based on extrap- Even when performed with great care, acousticanalyses olations from the sounds they produce. Hence, misana- may not lead to unambiguoussolutions. In developing his lyzed sonogramsconstitute noise in our data base.Let me ideas of how a syrinx might produce harmonic sounds, provide an example. Greenewalt(1968, Chap. 10) consideredthe fact that many I have chosen this example for several reasons. First, birds produce soundsin which the apparent fundamental the faults in acousticanalysis are irrelevant to the author’s and severallower harmonicsare missing.He hypothesized main interest. Thus, the paper is typical of many papers that these were suppressedby proximity of the vibrating that useacoustic data but are not concernedprimarily with membrane to the opposite trachea-bronchialwall. As one acousticanalysis. Second, the author recognizesthe lim- example, he presenteda detailed analysisof the Scold call