Empirical Musicology Review Vol. 5, No. 1, 2010 10 Fixed Average Spectra of Orchestral Instrument Tones JOSEPH PLAZAK Ohio State University DAVID HURON Ohio State University BENJAMIN WILLIAMS Ohio State University ABSTRACT: The fixed spectrum for an average orchestral instrument tone is presented based on spectral data from the Sandell Harmonic Archive (SHARC). This database contains non-time-variant spectral analyses for 1,338 recorded instrument tones from 23 Western instruments ranging from contrabassoon to piccolo. From these spectral analyses, a grand average was calculated, providing what might be considered an average non-time-variant harmonic spectrum. Each of these tones represents the average of all instruments in the SHARC database capable of producing that pitch. These latter tones better represent common spectral changes with respect to pitch register, and might be regarded as an “average instrument.” Although several caveats apply, an average harmonic tone or instrument may prove useful in analytic and modeling studies. In addition, for perceptual experiments in which non-time-variant stimuli are needed, an average harmonic spectrum may prove to be more ecologically appropriate than common technical waveforms, such as sine tones or pulse trains. Synthesized average tones are available via the web. Submitted 2010 January 22; accepted 2010 February 28. KEYWORDS: timbre, SHARC, controlled stimuli MOST modeling studies and experimental research in music perception involve the presentation or input of auditory stimuli. In many cases, researchers have aimed to employ highly controlled stimuli that allow other researchers to precisely replicate the procedure. For example, many experimental and modeling studies have employed standard technical waveforms, such as sine tones or pulse trains. In other cases, researchers have identified a particular spectral recipe, such as 10 equally-weighted harmonics (e.g. Plomp & Levelt, 1965; Huron & Sellmer, 1992). At the same time, researchers recognize the importance of using sounds that better approximate the sorts of sounds encountered in common listening situations. Many perceptual studies make use of commercial sound recordings, or experiment-specific recorded examples, played either on acoustical instruments or using MIDI devices. The choice of musically-pertinent stimuli is often regarded as a dichotomy between “control” and “ecological validity.” With current technology it is possible to have both: stimuli can be produced that closely resemble musically appropriate sounds, yet are sufficiently well defined so as to permit replication by other researchers. As a potential tool for researchers, we present in this paper average harmonic spectra for the pitches B0-G7, and also a grand average spectrum from the sum of these pitches. For many applications, these highly replicable spectra will prove more ecologically valid than technical waveforms. For many phenomena, the choice of stimulus materials may prove unimportant. When different stimuli converge on the same results, we may infer that the specific timbres employed are inconsequential. However, in many other cases, the choice of timbres can prove critical. The results for simple tones (such as sine waves) may differ from results using complex tones (such as recorded piano music). One possible approach to creating an ecologically useful musical stimulus is to identify an “average” musical sound. In this brief report, we describe such an effort. The notion of an average sound raises a host of questions related to the population of sounds for which some sound purports to be the average. An average value may not be representative. For example,