The Spectral Slope

spectral-slope spectral-tilt source-spectrum glottal-flow voice-quality acoustics
Last updated: 2025-02-07

The Spectral Slope

Spectral slope (also called spectral tilt or spectral roll-off) describes the rate at which acoustic energy decreases with increasing frequency in the spectrum. This fundamental characteristic of voice acoustics reflects the shape of the glottal flow waveform and provides valuable information about phonation type, vocal effort, and voice quality. Understanding spectral slope is essential for both theoretical models of voice production and clinical assessment of voice disorders.

Definition and Measurement

Spectral slope quantifies the general trend of amplitude decrease across the frequency range of a spectrum.

Conceptual Definition

In most voiced sounds, harmonic amplitudes progressively decrease as frequency increases. The spectral slope describes this decline:

Steep Negative Slope: Rapid decrease in amplitude with increasing frequency; energy concentrated in lower harmonics.

Shallow Negative Slope: Gradual decrease; substantial energy in higher harmonics.

Positive Slope: Unusual condition where higher frequencies have greater amplitude than lower frequencies (rare in normal phonation).

The slope reflects the overall “brightness” or “darkness” of the voice timbre—steeper slopes sound darker or more mellow, while shallower slopes sound brighter or more brilliant.

Quantification Methods

Several methods quantify spectral slope:

Regression-Based Slope: Fit a straight line (linear regression) to spectral data (typically in dB) vs. frequency:

Amplitude(dB) = Slope × Frequency + Intercept

The slope coefficient (typically negative) indicates dB change per Hz or per octave.

Harmonic Amplitude Ratios: Compare specific harmonic amplitudes:

H1-H2 = Amplitude(f₀) - Amplitude(2f₀)  [in dB]

H1-H2: Difference between first and second harmonic amplitudes. Positive values indicate declining slope; larger values indicate steeper slope.

H1-A1: Difference between first harmonic and amplitude of first formant region. Compensates for formant effects.

H1-A3: Difference between first harmonic and amplitude of third formant region. Sensitive to different spectral characteristics.

Spectral Moments: Statistical measures of spectral distribution:

  • Spectral centroid: Center of gravity of spectrum (Hz)
  • Spectral spread: Standard deviation around centroid
  • Higher centroid indicates shallower slope (energy at higher frequencies)

dB/Octave Measure: Amplitude decrease per frequency doubling:

Slope = [Amplitude(2f) - Amplitude(f)] / log₂(2f/f)

Typical voice source: approximately -12 dB/octave

Spectral slope examples Figure 5.6: Comparison of spectra with different spectral slopes, showing steep slope (top) with energy concentrated in low frequencies vs. shallow slope (bottom) with substantial high-frequency energy.

Physical Basis: Glottal Flow Waveform

Spectral slope directly reflects the time-domain shape of the glottal volume velocity waveform.

The Source Spectrum

The glottal source spectrum—the frequency content of airflow through the glottis—determines the initial spectral slope before vocal tract filtering.

Typical Characteristics:

  • Fundamental at F₀ with maximum amplitude
  • Harmonics at integer multiples (2F₀, 3F₀, …)
  • Progressive amplitude decline with increasing frequency
  • Slope approximately -12 dB/octave (amplitude decreases by factor of 4 per octave)

This -12 dB/octave baseline slope corresponds to a roughly triangular or differentiated pulse waveform shape in the time domain.

Waveform Shape and Spectral Slope

The relationship between time-domain waveform shape and spectral slope follows from Fourier analysis principles:

Sharp Discontinuities (abrupt changes in flow):

  • Produce energy at high frequencies
  • Result in shallower spectral slope
  • More high-frequency harmonics
  • Examples: abrupt glottal closure, short closing phase

Smooth, Gradual Changes:

  • Produce less high-frequency energy
  • Result in steeper spectral slope
  • Fewer strong high-frequency harmonics
  • Examples: gradual closure, incomplete closure

Mathematical Relationship: Each differentiation in the time domain corresponds to multiplication by frequency (f) in the frequency domain, adding approximately -6 dB/octave to the slope.

Glottal Pulse Parameters

Specific features of the glottal cycle affect spectral slope:

Closing Phase Duration:

  • Short closing phase: Abrupt flow cutoff → shallower slope (more high frequencies)
  • Long closing phase: Gradual flow reduction → steeper slope (less high frequencies)

Open Quotient (OQ): Ratio of open phase duration to total cycle period:

  • Low OQ (short open phase): Sharper pulses → shallower slope
  • High OQ (long open phase): Broader pulses → steeper slope

Glottal Closure Pattern:

  • Complete closure: Sharp flow discontinuity → shallower slope, strong high frequencies
  • Incomplete closure: Continuous DC flow component → steeper slope, weaker high frequencies

Pulse Skewness: Asymmetry between opening and closing phases:

  • Rapid closing, slow opening: Enhanced high frequencies
  • Slow closing, rapid opening: Reduced high frequencies

Phonation Types and Spectral Slope

Different phonation types produce characteristic spectral slopes reflecting their distinct glottal configurations.

Characteristics:

  • Complete glottal closure during closed phase
  • Moderate open quotient (0.4-0.6)
  • Relatively abrupt closure
  • Balanced opening and closing phases

Spectral Slope:

  • Approximately -10 to -12 dB/octave
  • H1-H2 typically 0-10 dB
  • Clear harmonics extending to 4-5 kHz
  • Moderate high-frequency energy

This balanced slope provides optimal acoustic energy distribution for speech intelligibility and vocal efficiency.

Breathy Phonation

Characteristics:

  • Incomplete glottal closure
  • Continuous airflow throughout cycle
  • High open quotient
  • Gradual flow variations

Spectral Slope:

  • Steeper than modal: -15 to -20 dB/octave or more
  • H1-H2 often >10 dB (H1 notably stronger)
  • Rapid high-frequency roll-off
  • Increased spectral noise between harmonics
  • Reduced overall harmonic amplitude

Perceptual Correlate: Breathy, soft, airy voice quality with reduced loudness and projection.

Clinical Significance: May indicate incomplete glottal closure from paresis, bowing, or hyperadduction disorders.

Pressed/Strained Phonation

Characteristics:

  • Forceful glottal closure
  • Extended closed phase (low open quotient)
  • Abrupt flow cutoff
  • High subglottal pressure

Spectral Slope:

  • Shallower than modal: -6 to -10 dB/octave
  • H1-H2 often <0 dB or negative (H2 stronger than H1)
  • Substantial high-frequency energy
  • Strong harmonics extending beyond 5 kHz
  • Possible subharmonics or nonlinear features

Perceptual Correlate: Harsh, strained, pressed voice quality; loud and projected but effortful.

Clinical Significance: May indicate hyperfunctional voice production, muscle tension dysphonia, or compensatory strategies.

Flow Phonation

Characteristics:

  • Balanced glottal contact
  • Efficient flow modulation
  • Optimal closing velocity
  • Typical of trained singers

Spectral Slope:

  • Moderate: approximately -10 to -12 dB/octave
  • Well-defined harmonics
  • Adequate high-frequency energy for resonance
  • Clean harmonic structure
  • Singer’s formant enhancement (3-4 kHz region)

Perceptual Correlate: Resonant, efficient, aesthetically pleasing voice with good projection and minimal effort.

Vocal Tract Effects on Spectral Slope

While glottal source characteristics primarily determine spectral slope, vocal tract resonances modify the observed output spectrum.

Source-Filter Separation

According to source-filter theory:

Output Spectrum = Source Spectrum × Vocal Tract Transfer Function

In logarithmic units (dB):

Output (dB) = Source (dB) + Filter (dB)

Source: Provides initial spectral slope from glottal flow.

Filter: Adds formant peaks and valleys, modifying local slope.

Output: Combined effect observed at lips.

Formant Effects

Formant Peaks: Local regions where vocal tract resonances add energy, creating local positive slopes in those frequency ranges.

Inter-Formant Valleys: Regions between formants where slope may be steeper than source slope.

Overall Trend: Global spectral slope still reflects source characteristics, but local deviations reflect formant structure.

To isolate source slope, researchers use:

  • Inverse filtering: Remove vocal tract effects computationally
  • Harmonic measurement in formant valleys: Measure where formants have minimal effect
  • LPC residual analysis: Extract excitation signal after removing vocal tract model

Lip Radiation Effect

Sound radiation from the mouth opening adds approximately +6 dB/octave boost:

Physical Basis: For wavelengths large compared to mouth opening (lower frequencies), mouth radiates omnidirectionally with reduced efficiency. For shorter wavelengths (higher frequencies), mouth acts as directional radiator with increased efficiency.

Net Effect: Partially compensates source slope. If source has -12 dB/octave and radiation adds +6 dB/octave, the radiated sound has approximately -6 dB/octave net slope.

This radiation characteristic is relatively constant across speakers and vocal configurations, so observed differences in output slope primarily reflect source differences.

Clinical and Diagnostic Applications

Spectral slope provides objective measures for clinical voice assessment.

Normal vs. Disordered Voice

Normal Voice:

  • Spectral slope within typical range (-10 to -12 dB/octave)
  • H1-H2 values consistent with phonation type
  • Clear harmonic structure throughout spectrum
  • Appropriate balance of low and high frequencies

Disordered Voice: Abnormal spectral slopes indicate specific pathophysiologies:

Steeper Slope (excessive high-frequency attenuation):

  • Incomplete glottal closure (paresis, bowing)
  • Reduced vocal fold tension
  • Edema or mass lesions
  • Breathy or weak voice quality

Shallower Slope (excessive high-frequency energy):

  • Hyperfunctional phonation
  • Muscle tension dysphonia
  • Excessive vocal effort
  • Harsh or strained quality

Acoustic Analysis Protocols

Clinical voice laboratories routinely measure:

H1-H2: Primary indicator of spectral slope H1-A1, H1-A3: Formant-corrected slope measures Spectral tilt: Overall slope across frequency range Cepstral peak prominence (CPP): Related to harmonic definition and slope regularity

These measures complement other acoustic parameters (jitter, shimmer, HNR) in comprehensive voice assessment.

Treatment Monitoring

Spectral slope changes track treatment effectiveness:

Voice Therapy: Successful therapy often normalizes spectral slope toward modal values.

Surgical Intervention: Medialization procedures reducing glottal gap typically decrease spectral slope (increase high-frequency energy).

Botulinum Toxin Injection: For spasmodic dysphonia may alter spectral slope depending on dose and injection site.

Serial measurements document objective changes corresponding to perceived voice quality improvements.

Spectral Slope in Singing and Professional Voice

Professional voice users manipulate spectral slope for acoustic and aesthetic effects.

Operatic and Classical Singing

Singer’s Formant: Trained singers develop enhanced energy in 2.5-4 kHz region, creating local spectral prominence.

Mechanism: Clustering of formants 3, 4, and 5 through specific vocal tract adjustments (lowered larynx, expanded pharynx).

Effect on Slope: Creates local interruption in overall decline, adding “ring” or “brilliance” to voice.

Function: Enables voice to project over orchestral accompaniment despite lower overall intensity.

Belt and Contemporary Commercial Music (CCM)

Belt Voice: High-intensity singing in chest register at high pitches.

Spectral Characteristics:

  • Relatively shallow slope compared to classical technique
  • Strong high-frequency harmonics
  • Enhanced energy in 2-3 kHz region (different from singer’s formant)
  • H1-H2 often near zero or negative

Acoustic Strategy: Maximizes acoustic power and penetration for amplified performance contexts.

Pedagogical Applications

Voice teachers use spectral characteristics (assessed perceptually or with real-time visual feedback) to guide students toward desired timbres:

“Darker” Tone: Encourage steeper slope through relaxed phonation, lower larynx position.

“Brighter” Tone: Encourage shallower slope through firmer closure, forward resonance focus.

Timbre Variety: Train ability to modify spectral slope intentionally for expressive purposes.

Measurement Considerations and Challenges

Accurate spectral slope measurement requires attention to methodological details.

Signal Selection

Sustained Vowels: Optimal for slope measurement; stable F₀ and formants.

Running Speech: More variable; consider long-term average spectrum (LTAS).

Pitch Effects: Standardize F₀ or measure across different pitches to assess range.

Analysis Parameters

Frequency Range: Define range for slope measurement (e.g., 0-5 kHz, F₀-5 kHz).

Formant Compensation: Decide whether to measure source slope (requires inverse filtering) or output slope (includes formant effects).

Window Length: Balance frequency resolution (longer windows) vs. temporal resolution (shorter windows).

Individual Differences

Anatomical Variation: Vocal tract size and shape affect formant spacing and radiation, influencing observed slope.

Habitual Phonation Pattern: Individual speakers have characteristic slopes even within “normal” range.

Age and Sex: Males typically have slightly different slope distributions than females; age-related tissue changes may affect slope.

Normative data stratified by age and sex provide appropriate comparison standards.

Summary

Spectral slope describes the rate of amplitude decrease with increasing frequency in the acoustic spectrum, typically quantified in dB/octave or through harmonic amplitude ratios like H1-H2. This fundamental characteristic directly reflects the time-domain shape of the glottal flow waveform—abrupt flow changes produce shallower slopes with substantial high-frequency energy, while gradual flow changes produce steeper slopes with attenuated high frequencies.

Different phonation types exhibit characteristic spectral slopes: modal phonation shows moderate slopes around -10 to -12 dB/octave, breathy phonation shows steeper slopes with enhanced H1-H2 ratios, and pressed phonation shows shallower slopes with reduced or negative H1-H2 values. Vocal tract filtering modifies the source slope through formant resonances and lip radiation effects, though the global trend primarily reflects glottal source characteristics.

Clinical voice assessment uses spectral slope measures to identify pathological phonation patterns, with abnormally steep slopes suggesting incomplete glottal closure and shallow slopes indicating hyperfunctional phonation. Professional singers manipulate spectral slope through technique to achieve desired timbres, including the classical singer’s formant and contemporary belt characteristics. Accurate slope measurement requires consideration of analysis parameters, formant compensation methods, and individual anatomical differences.

Key Takeaways

  • ✅ Spectral slope quantifies the rate of amplitude decrease with increasing frequency, typically around -10 to -12 dB/octave for modal phonation
  • ✅ Slope directly reflects glottal flow waveform shape: abrupt closure produces shallower slopes, gradual closure produces steeper slopes
  • ✅ Breathy phonation shows steep slopes (H1-H2 >10 dB) while pressed phonation shows shallow slopes (H1-H2 <0 dB)
  • ✅ Vocal tract filtering adds formant peaks and lip radiation effects (+6 dB/octave) that modify source slope
  • ✅ Clinical assessment uses slope measures to identify incomplete glottal closure (steep) or hyperfunctional phonation (shallow)
  • ✅ Professional singers manipulate spectral slope for acoustic effects like singer’s formant (2.5-4 kHz) in classical technique
  • ✅ Accurate measurement requires attention to frequency range, formant compensation, and individual anatomical differences

Further Reading

  1. Hanson, H. M., & Chuang, E. S. (1999). Glottal characteristics of male speakers: Acoustic correlates and comparison with female data. Journal of the Acoustical Society of America, 106(2), 1064-1077.
  2. Titze, I. R. (2000). Principles of Voice Production (2nd ed.). Iowa City: National Center for Voice and Speech.
  3. Sundberg, J. (1987). The Science of the Singing Voice. DeKalb, IL: Northern Illinois University Press.
  4. Klatt, D. H., & Klatt, L. C. (1990). Analysis, synthesis, and perception of voice quality variations among female and male talkers. Journal of the Acoustical Society of America, 87(2), 820-857.
  5. Alku, P., Bäckström, T., & Vilkman, E. (2002). Normalized amplitude quotient for parametrization of the glottal flow. Journal of the Acoustical Society of America, 112(2), 701-710.

Connections across chapters

Practice all connections

Each prompt ties this topic to another chapter. Predict the outcome before you open the model answer.

Ch. 5 · The Spectral Slope (this page) Ch. 10 · Temporal Gap and Spectral Slope

Closure abruptness is what the ear calls a register

A singer moves from modal voice to falsetto on the same pitch. Predict what happens to the glottal flow waveform, the spectral slope and the abduction quotient, and explain why a listener hears a categorical change rather than a gradual one.

Think it through, then show a model answer

In modal voice the folds close firmly and abruptly with the body participating, so the flow pulse has a sharp falling edge, the spectrum is rich and its slope shallow, around −6 to −12 dB per octave. In falsetto only the cover vibrates against a stiff, elongated fold, closure is gentle or incomplete, the flow pulse is nearly sinusoidal, the slope steepens to −12 to −24 dB per octave and the abduction quotient (prephonatory gap over amplitude) rises from below 0.3 to above 0.5. The physical change is continuous, but the ear integrates high-frequency energy and sorts voices into categories, so a modest change in slope crosses a perceptual boundary and is heard as a switch of register, with the modal–falsetto boundary wider and more controllable than the pulse–modal one.

A good answer mentions

  • Abrupt closure gives a shallow slope; gentle closure gives a steep slope
  • Abduction quotient rises from under 0.3 (modal) to over 0.5 (falsetto)
  • Categorical perception turns a continuous acoustic change into a heard register switch

Ch. 5 · The Spectral Slope (this page) Ch. 9 · Radiation from the Mouth

A −12 dB/octave source, a +6 dB/octave mouth

A modal voice source falls off at about −12 dB per octave. Predict the slope of the sound measured a metre in front of the mouth (ignoring the formants), explain the mechanism, and use it to say why [a] is louder than [i] at the same effort.

Think it through, then show a model answer

Radiation from the mouth acts as a differentiator with a +6 dB per octave characteristic: low frequencies are largely reflected back at the lips, high frequencies escape efficiently. Combined with the −12 dB per octave source this leaves a net slope near −6 dB per octave before the formants are superimposed. The same mechanism explains vowel loudness: a wide mouth opening on [a] radiates more efficiently, especially at low frequencies, and places F₁ high where strong low harmonics can be amplified, whereas the narrow opening of [i] reflects more energy back into the tract. The radiation boost is also why singers gain so much from energy in the 2–4 kHz region.

A good answer mentions

  • Radiation adds about +6 dB/octave, giving a net slope of about −6 dB/octave
  • Low frequencies reflect at the lips; high frequencies radiate efficiently
  • A larger mouth opening radiates better, so [a] is louder than [i]

Ch. 1 · Vascularity and Innervation of the Larynx Ch. 5 · The Spectral Slope (this page) Ch. 9 · Dependence of Glottal Source Power on Adduction

From a nerve to a spectrum

A patient has a unilateral recurrent laryngeal nerve paralysis. Name the muscles that no longer work on that side, then predict the glottal closure, the open quotient, the spectral slope and the loudness of the voice, and say which aspect of pitch control survives and why.

Think it through, then show a model answer

The recurrent nerve supplies every intrinsic muscle except the cricothyroid, so the thyroarytenoid, lateral and posterior cricoarytenoids and interarytenoid on that side are lost and the fold sits immobile in a paramedian position. Closure is incomplete, so the open quotient is high (above 0.7), a steady leak of air flows through the gap, and the glottal pulse closes gradually rather than abruptly. That gives a steep spectral slope (H1–H2 well above 10 dB), added aspiration noise and low intensity: the breathy voice of paralysis. The cricothyroid is supplied by the superior laryngeal nerve and still works, so the patient can still lengthen the folds and change pitch, though with a weak, breathy tone.

A good answer mentions

  • The recurrent nerve carries all intrinsic muscles except the cricothyroid
  • Incomplete closure raises open quotient and adds steady leakage flow
  • Gradual closure gives a steep spectral slope, aspiration noise and low intensity; cricothyroid pitch control is spared
Back to Chapter 05