Speech-in-Noise Testing, PB-Max & Retrocochlear Rollover Diagnostics

Speech-in-Noise Testing, PB-Max & Retrocochlear Rollover Diagnostics

Pure-tone audiometry and quiet word recognition testing assess only basic auditory sensitivity and maximal speech resolution under idealized, anechoic conditions. However, patients rarely complain of communication breakdowns in quiet, isolated sound booths; their primary functional deficit occurs in dynamic, competitive acoustic environments featuring multi-talker babble and reverberation. Furthermore, quiet testing frequently demonstrates ceiling effects that fail to expose neural transmission bottlenecks characteristic of retrocochlear lesions. Suprathreshold Performance-Intensity functions, retrocochlear rollover quantification, and speech-in-noise test batteries provide the essential diagnostic bridge between laboratory audibility and real-world auditory disability.

---

Performance-Intensity (PI) Functions for Speech

A Performance-Intensity (PI) function is a graphical plot illustrating a patient's speech recognition accuracy (percentage correct on the ordinate) across systematic increments of speech presentation level in $\text{dB HL}$ or $\text{dB SL}$ (on the abscissa).

Psychometric Slopes: Spondees vs. Monosyllables

• PI-Spondee Function: As noted in basic speech audiometry, spondees contain high lexical redundancy, balanced stress, and dual acoustic syllables. The psychometric slope of the PI-spondee function is exceptionally steep, averaging $10\%$ to $12\%$ per decibel near threshold. It rises rapidly from $0\%$ to $100\%$ over a narrow span of $10-12\,\text{dB}$.
• PI-PB Function (Phonetically Balanced Monosyllables): Monosyllabic words (NU-6, CID W-22) lack linguistic redundancy. The slope of the PI-PB function in normal-hearing individuals is substantially shallower, averaging $4\%$ to $5\%$ per decibel. Normal listeners require approximately $25-30\,\text{dB}$ above their SRT (i.e., $+25$ to $+30\,\text{dB SL}$) to attain maximum discrimination, reaching a $100\%$ ceiling that plateaus stably up to high presentation levels ($90-100\,\text{dB HL}$).

Morphological Profiles Across Auditory Pathologies

The morphology of the PI-PB function serves as an invaluable differential diagnostic indicator across the peripheral and central auditory pathways:

1. Conductive Hearing Loss: The middle ear attenuation acts as a simple mechanical volume control. The entire PI-PB function maintains its normal steep slope ($4-5\%/\text{dB}$) and reaches a normal ceiling ($96-100\%$), but the curve is shifted horizontally to the right by an amount precisely equal to the air-bone gap.
2. Cochlear (Sensorineural) Hearing Loss: Damage to the active cochlear amplifier (outer hair cell loss) and elevated basilar membrane damping cause two characteristic alterations:
   • The slope of the curve becomes shallower ($2-3\%/\text{dB}$).
   • The maximum score (PB-Max) is depressed (typically between $50\%$ and $90\%$ depending on the severity of inner hair cell loss and high-frequency configuration).
   • At high presentation levels ($85-100\,\text{dB HL}$), the curve reaches a plateau, occasionally showing a slight, physiologically benign drop ($< 10-15\%$) due to mechanical distortion and recruitment.
3. Retrocochlear (Eighth Nerve) Pathology: Pathologies affecting the acoustic nerve (cranial nerve VIII)—such as vestibular schwannomas (acoustic neuromas), cerebellopontine angle meningiomas, or auditory neuropathy spectrum disorder—produce a catastrophic breakdown in speech processing at high intensities known as rollover.

---

PB-Max & The Retrocochlear Rollover Phenomenon

PB-Max is defined as the absolute maximum word recognition percentage achieved by a patient at any presentation level across their individual Performance-Intensity function. It represents the patient's ultimate speech recognition potential when audibility is maximized.

Neurophysiological Mechanism of Rollover

In normal auditory nerve fibers, increasing stimulus intensity accelerates neural firing rates and recruits adjacent fibers up to physiological saturation. However, when an acoustic neuroma compresses the eighth cranial nerve, or when demyelinating disease impairs axonal conduction:

• The compressed, demyelinated fibers cannot sustain high-frequency action potential generation.
• Under high-intensity acoustic drive, neural refractoriness increases, synaptic transmitter stores at the inner hair cell ribbon synapse become depleted, and neural conduction blocks occur.
• Crucially, the spatial and temporal phase-locking of the neural volley is obliterated. The central nervous system receives a severely desynchronized, degraded neural discharge, resulting in a paradoxical, dramatic collapse in speech intelligibility as the stimulus is made louder.

The Rollover Index (RI) Formula

To objectively quantify the severity of the rollover phenomenon and differentiate benign cochlear hearing loss from retrocochlear lesions, James Jerger and Susan Jerger (1971) formulated the Rollover Index (RI): $\text{RI} = \frac{PB_{\text{max}} - PB_{\text{min}}}{PB_{\text{max}}}$ where:

• $PB_{\text{max}}$: The highest percentage correct score obtained across the entire PI function.
• $PB_{\text{min}}$: The lowest percentage correct score obtained at a presentation level higher than the level that produced $PB_{\text{max}}$.

Diagnostic Thresholds for Retrocochlear Pathology

Diagnostic criteria for the Rollover Index vary with the word list, because harder lists produce lower PB-Max values:

• PAL PB-50 Word Lists (original criterion): $\text{RI} > 0.45 \implies \text{Suggests Retrocochlear Pathology}$ (Jerger and Jerger, 1971; Dirks et al., 1977).
• NU-6 Word Lists (common clinical lists): a lower criterion of roughly $\text{RI} \ge 0.25$ (Bess, Josey, and Humes, 1979) to $0.35$ (Meyer and Mishler, 1985) is used.

Rollover is a screening sign, not a diagnosis: a positive result leads to medical referral and MRI.

Step-by-Step Clinical Calculation Example

A 52-year-old patient presents with asymmetric sensorineural hearing loss and unilateral left-sided tinnitus. The audiologist administers NU-6 word lists across multiple intensity levels to evaluate the left ear's PI function:

• $60\,\text{dB HL}: 56\%$
• $75\,\text{dB HL}: 80\%$
• $85\,\text{dB HL}: 82\% \implies PB_{\text{max}}$
• $95\,\text{dB HL}: 52\%$
• $105\,\text{dB HL}: 36\% \implies PB_{\text{min}}$

Applying the Rollover Index formula: $\text{RI} = \frac{0.82 - 0.36}{0.82} = \frac{0.46}{0.82} \approx 0.561$ Clinical Interpretation: Because the calculated Rollover Index of $0.56$ exceeds both the NU-6 criterion (about $0.25$ to $0.35$) and the stricter PB-50 criterion of $0.45$, this represents a positive rollover test. The patient must be immediately referred to neurotology for diagnostic contrast-enhanced magnetic resonance imaging (MRI) of the internal auditory canals and cerebellopontine angle.

---

The Clinical Rationale for Speech-in-Noise (SIN) Testing

For decades, clinical audiometry relied almost exclusively on pure-tone audiograms and quiet word recognition scores. This diagnostic paradigm suffers from severe clinical blind spots:

1. The "Hidden Hearing Loss" Conundrum: Many patients present with substantial, debilitating complaints of understanding conversation in group settings, restaurants, and meetings, yet exhibit normal or near-normal pure-tone audiograms ($0-20\,\text{dB HL}$). Animal and human temporal bone studies demonstrate cochlear synaptopathy—loss of synaptic connections between inner hair cells and low-spontaneous-rate, high-threshold auditory nerve fibers. These low-spontaneous-rate fibers are essential for coding speech in background noise but contribute negligibly to detection of soft pure tones in quiet.
2. Ceiling Effects of Quiet Testing: Monosyllabic words presented at $+30-40\,\text{dB SL}$ in quiet fail to tax auditory processing. Patients with moderate sensory loss frequently achieve $90\%$ scores in quiet booths, misleading clinicians and patients regarding the true extent of their disability.
3. Signal-to-Noise Ratio (SNR) Dynamics: In real-world listening, communication success is dictated not by absolute decibel level, but by the Signal-to-Noise Ratio (SNR): $\text{SNR} = \text{Target Speech Level (dB)} - \text{Competing Noise Level (dB)}$ Normal-hearing listeners can comprehend connected speech at negative or zero SNRs (e.g., $0\,\text{dB}$ to $-3\,\text{dB}$ SNR). Individuals with sensorineural hearing loss require a significantly more positive SNR ($+5\,\text{dB}$ to $+15\,\text{dB}$) to achieve equal comprehension due to widened cochlear filters and degraded temporal resolution.

---

Standardized Speech-in-Noise Evaluation Protocols

1. QuickSIN (Quick Speech-in-Noise Test)

Developed by Killion, Niquette, Gudmundsen, Revit, and Banerjee (2004), the QuickSIN is the most widely adopted clinical test for quantifying SNR loss.

• Stimuli & Structure: Employs low-context sentences formulated by the Institute of Electrical and Electronics Engineers (IEEE). Sentences feature low semantic predictability (e.g., "The birch canoe slid on the smooth planks"), forcing listeners to rely purely on acoustic-phonetic processing. Each sentence contains 5 scored keywords.
• Competing Noise: Four-talker babble (three female voices and one male voice), creating a realistic, energetically and informational masking background.
• Presentation Methodology: A complete test list consists of 6 sentences presented at pre-recorded, decreasing SNRs in $5\,\text{dB}$ decrements: $+25\,\text{dB}, +20\,\text{dB}, +15\,\text{dB}, +10\,\text{dB}, +5\,\text{dB}, \text{and } 0\,\text{dB SNR}$. The speech is fixed at a comfortable suprathreshold level (typically $70\,\text{dB HL}$ for pure-tone averages $\le 45\,\text{dB HL}$, or "Loud but OK" for greater hearing losses).
• SNR Loss Mathematical Formulation: $\text{SNR Loss} = 25.5 - \text{Total Words Correct}$ Mathematical Derivation: By the Spearman-Kärber method, the SNR for 50% correct (SNR-50) equals $27.5$ (the $25\,\text{dB}$ starting SNR plus half the $5\,\text{dB}$ step) minus the total words correct. Normal-hearing listeners have an SNR-50 of about $2\,\text{dB}$, so SNR loss $= (27.5 - \text{words correct}) - 2 = 25.5 - \text{words correct}$.

Critical Test-Retest Rule: A single QuickSIN list has a $95\%$ critical difference of about $\pm 2.7\,\text{dB}$. To achieve high clinical precision, an audiologist must average at least two lists (reducing the $95\%$ critical difference to $\pm 1.9\,\text{dB}$) or three lists for medicolegal or research applications.

2. BKB-SIN (Bamford-Kowal-Bench Speech-in-Noise)

• Target Population: Specifically designed for pediatric patients (developmental age $5\,\text{years}$ and older), cochlear implant candidates, and adults with severe cognitive impairment or limited English proficiency who cannot process complex IEEE sentences.
• Stimuli: Employs syntactically simple BKB sentences (1st to 2nd-grade reading level; e.g., "The clown had a funny face") with 3 scored keywords per sentence.
• Acoustic Progression: Sentences are presented in four-talker babble from $+21\,\text{dB}$ down to $-6\,\text{dB SNR}$ in $3\,\text{dB}$ steps.
• Scoring: $\text{SNR-50} = 23.5 - \text{Total Key Words Correct}$ SNR loss is then the listener's SNR-50 minus the age-appropriate normal SNR-50 published in the test manual.

3. WIN (Words in Noise Test)

• Stimuli: Monosyllabic words from the Northwestern University Auditory Test No. 6 (NU-6) presented in multi-talker babble (Wilson et al.).
• Methodology: Assesses word recognition in noise at 7 discrete SNRs ($+24\,\text{dB}$ to $0\,\text{dB}$ in $4\,\text{dB}$ steps). Because it utilizes isolated monosyllables, the WIN strips away all sentence-level syntactic and contextual cues, directly testing the auditory periphery's ability to extract lexical tokens from competing noise.

4. HINT (Hearing in Noise Test)

• Adaptive Methodology: Developed by Nilsson, Soli, and Sullivan (1994). Unlike fixed-level tests, the HINT is an adaptive test that measures the Speech Reception Threshold in Noise (SRT-N)—the precise SNR in $\text{dB}$ required for the listener to correctly repeat $50\%$ of presented sentences.
• Stimulus & Noise Setup: Continuous speech-spectrum noise is delivered at a fixed sound pressure level (typically $65\,\text{dB A}$). The speech level increases or decreases in $2-4\,\text{dB}$ adaptive steps based on the accuracy of the preceding sentence.
• Spatial Release from Masking: HINT evaluates spatial hearing by presenting sentences under three soundfield speaker configurations: (1) Noise Front ($0^\circ$), (2) Noise Right ($90^\circ$), and (3) Noise Left ($270^\circ$). A Composite Threshold is computed. Comparing frontal noise to asymmetric noise conditions objectively measures the patient's head shadow advantage and central binaural squelch (spatial release from masking, which typically provides a $3-5\,\text{dB}$ benefit in healthy binaural systems).