Can Everyone Hear Your Calls? Why Open Office Speech Privacy Fails

The Statistic That Should Worry Every Facilities Manager

In a 2021 Leesman Index survey of 750,000 office workers across 5,000 workplaces worldwide, acoustic privacy was rated as "important" or "very important" by 78% of respondents — yet only 29% reported that their workplace provided adequate acoustic privacy. This 49-percentage-point gap was the largest satisfaction deficit of any workplace feature measured, exceeding dissatisfaction with temperature, air quality, natural light, and even desk space.

The consequence is measurable. Banbury and Berry (2005) demonstrated a 15–28% reduction in performance on tasks requiring concentration when workers were exposed to intelligible background speech, compared to unintelligible speech or silence. The key word is "intelligible" — it is not the volume of speech that causes distraction, but its intelligibility. The brain is hardwired to process speech that it can understand, and this processing consumes cognitive resources whether the listener wants it to or not.

The metric that quantifies this is the Speech Transmission Index — STI. If the STI between a talker and a listener exceeds 0.50, the speech is intelligible enough to distract. WELL v2 Feature 74 Part 3 sets this as the threshold: STI must be below 0.50 in open plan work areas. The majority of offices fail.

How STI Determines Speech Privacy

The Intelligibility-Distraction Threshold

STI measures the degree to which the temporal modulations of speech are preserved between the talker and the listener, on a scale from 0 (no modulation preserved — speech is completely unintelligible) to 1 (perfect modulation preservation — speech is perfectly clear). The calculation methodology is defined in IEC 60268-16:2020 §4, and the relationship between STI and intelligibility is well-established:

For speech privacy, the classification scale is inverted relative to its use in auditoria and classrooms. In a lecture hall, "excellent" STI is the goal. In an open plan office, "poor" STI is the goal. An STI of 0.35 at a neighbouring workstation means that speech is audible as a vague murmur but cannot be decoded into words — the brain does not attempt to process it, and cognitive distraction is minimal.

The Three Factors That Determine Office STI

STI between two positions in an open office is determined by three physical factors, commonly known as the ABC rule:

A — Absorption: The amount and distribution of sound-absorbing material in the space, primarily the ceiling. High absorption reduces the reverberant field, which means speech energy from the talker decays more rapidly with distance and fewer reflections reach the listener. The key metric is RT60 — target 0.4–0.6 seconds for open plan offices.

B — Barriers: Physical screens, partitions, and furniture between talker and listener that block the direct sound path. A desk screen of 1200–1400 mm height can reduce the direct sound level by 5–10 dB at seated ear height, provided the screen is solid (not perforated) and extends at least 400 mm beyond the seated user's head position on each side.

C — Cover (sound masking): A continuous, spectrally shaped background noise signal that raises the ambient level from the typical office background of 30–35 dBA to 40–45 dBA. This additional 10 dBA of masking noise reduces the signal-to-noise ratio at the listener position, which directly reduces STI.

Each factor addresses a different component of the STI equation. Absorption reduces the reverberant speech level. Barriers reduce the direct speech level. Masking raises the noise floor. All three must be present for STI to drop below 0.50 at typical office workstation spacings of 3–5 metres.

The Physics of Speech Privacy Failure

Why Most Offices Fail

Consider a typical modern open plan office with the following characteristics:

• Ceiling: Suspended mineral fibre acoustic tile, NRC 0.85
• Floor: Commercial carpet tile
• Desk screens: 1100 mm high fabric-covered steel
• Background noise: 32 dBA (HVAC at low speed, evening measurement)
• Workstation spacing: 3.2 m centre to centre (typical 1600 mm desk depth x 2)

Step 1: Direct sound level at 3.2 m

Normal speech at 1 metre is approximately 60 dBA. The 1100 mm screen provides approximately 5 dB insertion loss at seated ear height for the A-weighted speech spectrum (based on empirical data from ISO 3382-3:2012 Annex A). Distance attenuation from 1 m to 3.2 m in a semi-reverberant space is approximately 7 dB (less than the free-field 10 dB because the ceiling and floor provide supporting reflections).

Direct + diffracted speech level at listener: 60 - 7 - 5 = 48 dBA

Step 2: Reverberant speech level

With NRC 0.85 ceiling tiles across the full ceiling area and RT60 of approximately 0.5 seconds, the reverberant field contributes an additional component. For an open plan space where the reverberant field is not fully diffuse (it is predominantly a 2D field between floor and ceiling), the reverberant contribution at 3.2 m can be estimated at approximately 4 dB below the direct component: 44 dBA.

Step 3: Combined speech level at listener

The energetic sum of direct (48 dBA) and reverberant (44 dBA) components:

L_combined = 10 x log10(10^4.8 + 10^4.4) = 10 x log10(63,096 + 25,119) = 10 x log10(88,215) = 49.5 dBA

Step 4: Signal-to-noise ratio

SNR = 49.5 - 32 = 17.5 dB

Step 5: Estimated STI

At this SNR and RT60, the STI can be estimated from the modulation transfer function. For a reverberant room with T60 = 0.5 s and SNR = 17.5 dB, the STI is approximately 0.62.

This is well above the 0.50 threshold. Speech from the neighbouring workstation is clearly intelligible. The office has good absorption, reasonable screens, but inadequate background noise — and it fails the WELL v2 Feature 74 Part 3 requirement.

The Missing Element: Background Noise Is Too Low

The office achieves excellent RT60, decent screening, but a background noise level of only 32 dBA. This is too quiet. In a very quiet office, even attenuated speech remains intelligible because the signal-to-noise ratio remains high.

If the background noise were raised to 42 dBA (through a sound masking system), the calculation changes:

SNR = 49.5 - 42 = 7.5 dB

At this SNR and the same RT60, the STI drops to approximately 0.44 — below the 0.50 threshold. Speech from the neighbouring workstation transitions from "clearly intelligible" to "audible murmur." The 10 dBA increase in background noise, achieved through a carefully tuned masking system, reduces STI by 0.18 and transforms the privacy performance from "fail" to "pass."

ISO 3382-3:2012 — The Open Plan Office Standard

ISO 3382-3:2012 provides the measurement framework for evaluating open plan office acoustics. It defines four parameters that together characterise the acoustic environment, as specified in §4:

D2,S — Spatial Decay Rate of Speech

The rate at which A-weighted speech sound pressure level decreases with distance doubling, measured in dB per distance doubling (dB/dd). This captures the combined effect of absorption and barriers on speech propagation.

• Target: D2,S >= 7 dB/dd per ISO 3382-3:2012 §4.2
• Typical poorly performing office: 3–4 dB/dd
• Well-performing office with ABC treatment: 8–12 dB/dd

Lp,A,S,4m — Speech Level at 4 Metres

The A-weighted sound pressure level of normal speech measured at 4 metres from the talker. This is the single most direct measure of privacy — if speech at 4 metres is below the masking level, it will not be intelligible.

• Target: Lp,A,S,4m <= 48 dBA per ISO 3382-3:2012 §4.3
• Typical poorly performing office: 52–55 dBA
• Well-performing office: 44–48 dBA

rD — Distraction Distance

The distance from the talker at which STI drops below 0.50. Beyond this distance, speech transitions from "distracting" to "non-distracting."

• Target: rD <= 5 metres
• Typical poorly performing office: rD = 10–15 metres (speech distracting across entire floor plate)
• Well-performing office: rD = 4–5 metres (only immediate neighbours distracted)

rP — Privacy Distance

The distance at which STI drops below 0.20. Beyond this distance, speech content cannot be reconstructed even with effort.

• Target: rP as low as possible, typically 8–12 m in well-designed offices
• Typical poorly performing office: rP = 15–25 metres

Worked Example: Achieving STI < 0.50 in an Open Plan Office

The Space

• Dimensions: 30 m x 20 m open plan floor plate
• Ceiling height: 2.7 m (suspended ceiling)
• Workstations: 80 desks in clusters of 4, 3.2 m spacing
• Floor area: 600 m²

Baseline Condition (Absorption Only, No Masking)

Improvement Step 1: Increase Screen Height to 1400 mm

Raising desk screens from 1100 mm to 1400 mm increases the insertion loss from approximately 5 dB to 8–10 dB at seated ear height. This is because the higher screen moves the diffraction edge further above the line of sight, increasing the path length difference between direct and diffracted paths.

Effect on STI at 3.2 m: reduces from 0.62 to approximately 0.55. Still above 0.50.

Improvement Step 2: Add Sound Masking at 42 dBA

A sound masking system installed above the suspended ceiling, consisting of loudspeakers on a 3 m x 3 m grid (approximately 67 speakers for 600 m²), producing a spectrally shaped broadband noise signal at 42 dBA (+/- 1 dB uniformity across the floor plate).

The masking spectrum should follow the STC-like contour recommended by ASTM E1573-21, with peak energy in the 200–500 Hz range (where speech energy is concentrated) and a gradual roll-off above 1 kHz (to avoid the "hissing" quality that occupants find annoying).

Effect on STI at 3.2 m with 1400 mm screens:

SNR = (48 dBA speech at listener) - (42 dBA masking) = 6 dB

At RT60 = 0.5 s and SNR = 6 dB, STI = approximately 0.40

This is below the 0.50 threshold. The combination of 1400 mm screens and 42 dBA masking achieves WELL v2 Feature 74 Part 3 compliance.

Summary of ABC Treatment Effects

The table demonstrates why all three elements of the ABC rule are necessary. Absorption alone achieves nothing. Barriers alone are insufficient. Masking alone creates an uncomfortably noisy environment. The combination of all three produces an acoustic environment where speech privacy is achieved at typical workstation distances.

Sound Masking: The Most Misunderstood Element

What Sound Masking Is

Sound masking is a system that generates a continuous, low-level broadband noise signal distributed uniformly throughout the workspace via loudspeakers installed above the ceiling. The noise is spectrally shaped to match the frequency profile of human speech — concentrating energy in the 200 Hz to 5 kHz range — so that it most effectively masks the speech frequencies that carry intelligibility.

The masking signal is not "white noise" (equal energy per Hz). White noise sounds harsh and hissing. Properly specified masking uses a "pink-ish" spectrum with additional roll-off above 2 kHz, producing a sound similar to gentle air conditioning. When correctly installed and tuned, occupants are not consciously aware of the masking — it blends with the ambient environment and sounds like "the building breathing."

What Sound Masking Is Not

Sound masking is frequently confused with two other concepts:

It is not noise cancellation. Active noise cancellation (ANC) uses an anti-phase signal to reduce a specific noise source. This works in headphones (small, enclosed volume) but is physically impossible to achieve across an open office floor plate. Sound masking does not cancel speech — it covers it.

It is not soundproofing. Soundproofing uses mass, isolation, and damping to prevent sound from transmitting between spaces. Sound masking does not reduce the speech level at the listener — it raises the noise floor to reduce the signal-to-noise ratio.

Specification and Commissioning

The masking system must be commissioned by a qualified acoustician. The commissioning process involves:

1. Level setting: Adjusting the overall masking level to 40–45 dBA at desk height. Too low (below 38 dBA) provides insufficient masking. Too high (above 48 dBA) is fatiguing and triggers its own Lombard effect.

1. Spectral shaping: Adjusting the frequency content to match the target spectrum. The 1/3-octave band levels should follow a smooth curve with peak energy at 250–500 Hz and a roll-off of approximately 3 dB per octave above 1 kHz.

1. Uniformity verification: Measuring the masking level at multiple positions across the floor plate to verify +/- 1 dB uniformity. Hot spots (where masking is too loud) create local discomfort. Dead spots (where masking is too quiet) create local privacy failures.

1. Ramping: Many installations ramp the masking level up gradually over 2–3 weeks, starting at 2–3 dB below the target and increasing by 1 dB per week. This allows occupants to acclimatise without conscious awareness.

Cost Analysis for a 600 m² Office

Against the productivity cost of speech distraction (15–28% performance reduction for knowledge workers earning an average of £45,000/year, affecting 80 workers):

Annual productivity loss = 80 x £45,000 x 0.20 (midpoint) = £720,000/year

Even if acoustic treatment recovers only half of this loss (a conservative assumption — Hongisto et al., 2016, measured 5–8% productivity improvement from sound masking alone), the annual benefit is £360,000 against a one-off cost of £31,000–£57,500. The payback period is approximately 5–8 weeks.

Common Mistakes

Mistake 1: Relying on Absorption Alone

Many office fit-outs specify premium acoustic ceiling tiles (NRC 0.85–0.95) and assume this solves the privacy problem. As the worked example demonstrates, even excellent ceiling absorption with RT60 of 0.5 seconds yields STI of 0.55–0.62 at typical desk spacings when background noise is low. Absorption is necessary but not sufficient.

Mistake 2: Screens That Are Too Short

A 900 mm desk screen — the most commonly specified height — provides virtually no acoustic benefit for seated users. The screen must extend at least 150 mm above seated ear height (approximately 1150 mm from the desk surface, or 1400 mm from the floor for a standard desk height). Screens below this height are furniture, not acoustic barriers.

Mistake 3: Masking That Is Too Loud

Some facilities managers, discovering the concept of sound masking, set the level to 48–50 dBA in an attempt to maximise privacy. This creates a new problem: the masking noise itself becomes fatiguing. Workers raise their voices to speak over the masking (the Lombard effect), which increases the speech level at neighbouring workstations, partially negating the privacy benefit. The optimal masking level is 40–45 dBA — loud enough to mask speech at typical distances but quiet enough to be unobtrusive.

Mistake 4: Ignoring Ceiling Void Paths

In offices with a suspended ceiling, sound can travel through the ceiling void (the space between the suspended ceiling and the structural slab above) and re-enter the workspace at a distant location. If the ceiling tiles have poor sound transmission loss (STC < 25), this "flanking path" bypasses the acoustic treatment in the workspace below. Sound masking speakers installed above the ceiling partially address this by filling the void with masking energy, but the ceiling tiles themselves must have adequate mass — a minimum surface density of 3.5 kg/m² is recommended for effective privacy.

WELL v2 Compliance Checklist

For projects pursuing WELL v2 Feature 74 Part 3 (Speech Privacy):

1. Measure background noise level at workstation positions (unoccupied, HVAC operating normally)
2. Calculate or measure STI from a representative talker position to neighbouring workstations
3. If STI > 0.50, determine which ABC element(s) are deficient
4. Specify treatment: ceiling absorption (A), screen height and density (B), masking system (C)
5. After installation, commission and verify: measure STI at representative positions to confirm < 0.50
6. Document results for WELL assessor submission — include STI measurement methodology per IEC 60268-16:2020 §5

Related Reading

• Open Plan Office Acoustic Design: The Complete Guide — the comprehensive reference covering all aspects of open plan acoustics
• WELL v2 Feature 74 Decoded: Every Acoustic Requirement — detailed breakdown of all three parts of Feature 74
• Speech Privacy Index (SPI) Explained — the ASTM E1130 privacy measurement framework