How Accurate Are ‘Speech Communication Method’-Generated Noise Levels?

In April 2019, the results of a Health and Safety Executive (HSE)-funded study were published in the International Journal of Audiology.

The research paper, ‘A simple method to estimate noise levels in the workplace based on self-reported speech communication effort in noise’[i] was co-authored by Professor of Audiology, Mark Lutman.

Of course, the University of Southampton Professor is widely renowned for his significant contribution to the ‘Guidelines on the diagnosis of noise-induced hearing loss for medicolegal purposes’ (Coles Lutman Buffin Guidelines), in 2000, and the ‘Guidelines for quantification of noise-induced hearing loss in a medicolegal context’ (Lutman Coles Buffin Guidelines), in 2015.

In his latest study, Lutman et al., identified that objective noise level data, recorded with sound level meters and personal dosimetry, is ‘seldom’ available in research studies or in personal injury claims. This places greater reliance on subjective, self-reported noise level data, which, at least in theory, does ‘not allow for accurate assessment of noise levels and cumulative noise exposure’.

The ‘speech communication method’ has been used, in past research, to estimate retrospective occupational noise levels. This technique involves participants describing the vocal effort required to communicate with a person at a typical conversational distance (1.2 metres) in a given occupational setting. Noise levels are then assigned to descriptions of communication ability.

As a result, the purpose of the 2019 study was to evaluate whether the ‘speech communication method’ provided noise levels that were as valid as objectively measured noise levels.

To do so, the study authors observed a group of participants, between the ages of 16 and 25 years, all of whom were exposed to occupational noise levels in excess of 85 dB(A) for at least 1-hour per day, in the course of the following job roles:

  • Pottery manufacturing;
  • Fettling;
  • Automotive assembly;
  • Paint shop work;
  • Shot blasting;
  • Welding;
  • Metal working;
  • Metal casting;
  • Food processing;
  • Window frame manufacturing;
  • Hydraulic pressing;
  • Petrochemical maintenance;
  • Pallet loading;
  • Pharmaceutical manufacturing;
  • Turbine engine manufacturing; and
  • Paint manufacturing.

Those with pre-existing hearing threshold levels above 30 dB, a sign of pathological hearing loss, were excluded from the study.

The participants were asked to rate the vocal effort required to maintain a conversation with a colleague, who was facing them at a 1.2 metre distance, above a normal level [<81 dB(A)]:

  • Raised voice [equating to 87 dB(A)];
  • Very loud voice [equating to 93 dB(A)]; or
  • Shouting [equating to 99 dB(A)].

This process was repeated for every main task within each workplace, where neither person had worn hearing protection, where the listener did not have a hearing impairment themselves, and where normal gesturing was used. Only speech communication abilities relating to current employment were compared with objectively measured noise levels. There were 168 self-reports in total.

Meanwhile, objective continuous noise levels [Leq dB(A)] for each working task were calculated with calibrated personal noise dosimetry badges, worn on the tops of the participants’ shoulders. There were 134 tasks performed in current workplaces, meaning that the 34 additional self-reports represented instances where participants worked on the same task in the same area of the workplace, i.e. they were duplicates.

When the objective and subjective measurements were compared, there was a symmetrical distribution of discrepancy, meaning that around 50% of the participants’ self-reports were overestimations of dosimetry and 50% were underestimations. Symmetrical distribution typically signifies a lack of bias.

What is more, 91% of the study group reported noise levels within ±6 dB of objectively measured levels and, of these participants, 56% reported noise levels within ±3 dB. This suggests that subjectively reported noise levels predicted actual noise levels with ‘good accuracy’.

In the remaining 9% of study group participants, i.e. those who provided estimated noise levels outside of dosimetry by more than ±6 dB, 6 individuals reported that they did not need to use a raised voice for tasks measuring up to 96 dB(A) – a ‘substantial underestimation’.

Nevertheless, 96% of the study group reported having to use a very loud voice to converse, inferring noise levels of 93 dB(A). In these cases, the mean difference between reported and measured noise levels was less than 3.2 dB.

The authors summarised the key findings as follows:

‘The results show that the method provides noise level estimates having a high probability (>90%) of being within 6 dB(A) of the actual noise level occurring within the workplace. Where noise levels are measured at ≤93 dB(A) the correspondence between estimated and measured noise levels is likely to be within 3 dB. Therefore, participants who reported the need to use a raised voice to hold a conversation with a colleague in the workplace at 1.2m away, which equates to 87 dB(A) on the speech communication table, were highly likely to be working in levels of noise exceeding 81 dB(A). This is above the lower exposure action value of 80 dB(A) in the CNWR, above which an employer must measure noise levels, then monitor and identify risks hearing of employees. Those who reported needing to use a very loud voice (corresponding to 93 dB(A)) would have a high probability of working in noise at or above the second action level of 85 dB(A). At this level employers must ensure that hearing protection is worn and reduce noise levels at source where possible’.

How are the study results likely to impact upon noise-induced hearing loss claims? The authors went on to theorise that:

‘Employee report of required speech communication effort in noise can be an effective method of retrospectively estimating noise levels within the workplace for use in retrospective epidemiological research. This may also be used for other studies where there is no access to dosimetry or formal noise surveys, as well as in medicolegal work where unbiased reports of communication effort are available.

Although this research may be contextually relevant to medical expert-drafted evidence in NIHL claims, self-reported claimant evidence, in the absence of noise surveys, is arguably most relevant to engineering evidence and discussion on breach of duty.

In the recent case of Mackenzie v Alcoa Manufacturing (GB) Ltd [2019] EWHC 149 (QB) (reported here), the High Court interpreted that the duty of employers to carry out noise assessments, under Regulation 4 of the Noise at Work Regulations (1989), was in force from 1970, when industry guidance was prescribed by non-binding publications: Noise and the Worker (1963) and The Code of Practice (1972).

The effect of this was to reverse the burden of proof from the claimant (proving excessive noise exposure) to the defendant (proving under-excessive noise exposure).

Plainly, the issue here, is that an automatic finding on breach, where 1970’s exposure is alleged, may rest on witness recollection of noise exposure that is not just subjective, but also historic. As HHJ Inglis, in the case of Parkes v Meridian Ltd [2007] EWHC B1 (QB), explained:

‘Descriptions of the need to raise voices are very subjective; most of the witnesses are describing events many years ago ... I do not think that the decision on noise levels ... can be affected by the lay witness evidence’.

This position was further cemented by Garnham J, in The National Union of Mineworkers v Organisation Internationale De L'Energie Et Des Mines [2019] EWHC 1359 (Comm), reported here:

‘I have regard to the common human capacity and tendency for a witness genuinely but mistakenly to recollect past events as having actually happened in the way in which the witness now with hindsight believes they would, or indeed should, have happened. In that respect I have also had regard to ...the unreliability of human memory’.

In summary, while Lutman et al., may have shown that ‘speech communication method’ is moderately accurate in documenting contemporaneous noise levels, more research is needed to assess whether anecdotal evidence, given decades after an alleged period of excessive noise exposure (or any exposure type for that matter), potentially by elderly claimants, should be admissible in legal practice.


[i] Ferguson MA et al., A simple method to estimate noise levels in the workplace based on self-reported speech communication effort in noise. Int J Audiol. 2019 Jul;58(7):450-453. <> accessed 8 July 2019.