Can sound quality be measured?
For engineers, the question seems perfectly reasonable. Modern acoustic analysis can measure sound pressure levels, frequency content, vibration, loudness, roughness, sharpness, tonal components, and countless other characteristics. Faced with such an abundance of data, it is tempting to assume that product sound quality can ultimately be reduced to a collection of numbers. If we can measure a sound accurately enough, surely we can determine whether it is good or bad.
During his online guest lecture for Edinburgh Napier University, David Bowen spent much of his time explaining why the answer is not nearly so simple. Across more than three decades working in acoustics, vibration, psychoacoustics, and product sound quality, Bowen has helped organisations understand how people respond to the sounds products make. Throughout a career spanning industrial research, consultancy, and product development, he has worked at the intersection of acoustics, psychoacoustics, engineering, and product design. Again and again, his examples pointed towards the same conclusion. Sound can be measured. Sound quality cannot.
This distinction formed the foundation of the lecture. Sound quality, Bowen argued, is not a property of a product. It is a response of people. A microphone does not experience annoyance. A sound level meter does not perceive quality. Only listeners do. Understanding product sound quality therefore requires understanding both the physical sound and the human beings who hear it. Difficulties emerge as soon as engineers attempt to connect measurements to human responses. Bowen illustrated this challenge through examples in which sounds with similar measured levels produced dramatically different subjective reactions. A pure tone, broadband noise, an organ note, or a piece of industrial machinery may all produce similar sound levels, yet listeners often describe them in very different ways. Some sounds are judged pleasant. Others are irritating. Some feel powerful. Others feel weak. Part of the difficulty lies in the way human hearing operates. Psychoacoustics has demonstrated repeatedly that listeners do not experience sound in a simple or linear fashion. Sensitivity varies across frequencies. Loudness does not increase proportionally with sound pressure. Perception depends not only upon what reaches the ears but also upon how the brain interprets it. Measuring the sound itself is only part of the problem.
Bowen illustrated this point through several examples that challenge common assumptions about listening. Human memory for loudness is surprisingly limited. When listeners hear two sounds separated by even a relatively short interval, their ability to compare levels accurately begins to deteriorate. Judgements become influenced by expectation, context, and interpretation rather than purely acoustic characteristics. Even when measurements are reliable, the perceptual processes through which listeners experience those sounds remain considerably more complex.
For decades, researchers attempted to bridge this gap through increasingly sophisticated metrics. If sound pressure level proved insufficient, perhaps loudness would provide a better predictor. If loudness proved inadequate, perhaps perceived noisiness, roughness, sharpness, or other psychoacoustic measures would help. Each new metric offered valuable insights, yet each also revealed new limitations. Bowen discussed how the arrival of jet aircraft exposed weaknesses in existing approaches to noise evaluation, prompting the development of measures intended to capture perceived noisiness more effectively. Those measures improved predictions in some contexts while proving less successful in others. Similar challenges emerged across industrial machinery, transportation systems, and consumer products. As soon as one perceptual factor appeared understood, another emerged. Listening proved stubbornly resistant to simple description.
Bowen’s career spans a period during which acoustics increasingly recognised that physical measurements alone could not explain human responses. Successive generations of psychoacoustic metrics attempted to narrow the gap between measurable sound and perceived quality. Each represented an improvement upon what came before, though none provided a complete solution. Human perception remained influenced by context, expectation, memory, meaning, and experience. The history of product sound quality therefore became, in part, a history of increasingly sophisticated attempts to understand how people listen. Similar problems emerge elsewhere. A piano recording played backwards retains many of its measurable characteristics, yet listeners immediately perceive something fundamentally different. What sounds like a piano becomes something closer to an organ. Human listeners detect meaningful changes that conventional measurements often struggle to explain. Again and again, perception proves more complicated than measurement.
If measurements alone cannot fully predict how people will respond, a difficult question follows. How should products be designed?
For Bowen, the answer lies in listening. Much of the lecture focused on sound quality jury testing, a methodology that places human listeners at the centre of the evaluation process. Rather than asking which sound measures best, researchers ask which sound people prefer, which sound communicates particular qualities, and which sound supports the intended experience of a product.
This creates an interesting tension. Engineers naturally seek measurements. Manufacturers want targets that can be specified, monitored, and improved. Product development processes favour quantities that can be compared and optimised. Yet listeners remain the ultimate judges of quality. No matter how sophisticated a measurement becomes, a product succeeds or fails according to how people experience it. Jury testing therefore emerged not as a rejection of engineering but as a recognition that engineering alone could not answer every question.
Carefully designed listening tests provide information that measurements alone cannot. This approach complements rather than replaces traditional acoustic analysis. Measurements help researchers understand what a product is doing acoustically. Listening tests help them understand how people respond. Product sound quality emerges through the relationship between these two perspectives. Designing listening tests of this kind is far from straightforward. Participants must be selected carefully. Stimuli need to be prepared consistently. Presentation order can influence responses. Questions must be designed in ways that avoid leading participants towards particular conclusions. Statistical analysis becomes essential if meaningful patterns are to emerge from the resulting data. Throughout the lecture, Bowen emphasised that listening tests require as much methodological care as any engineering measurement.
One particularly interesting aspect of this work involves the creation of what Bowen described as virtual products. Rather than constructing numerous physical prototypes, researchers can isolate individual sound components and manipulate them independently. Motor noise, airflow noise, pump sounds, valve sounds, and other elements can be adjusted before being recombined into new versions of the product. Listeners can then evaluate these alternatives, allowing researchers to explore how specific design decisions influence perceived quality without repeatedly redesigning the product itself.
One of the lecture’s most illuminating examples involved front-loading washing machines. Modern washing machines generate a wide range of sounds, including motor noise, water movement, pumping systems, valves, and the movement of clothes within the drum. Traditional noise control might focus simply on reducing these sounds wherever possible. Bowen’s research adopted a different approach. Rather than treating the machine as a single noise source, the different sounds produced during filling, washing, draining, and spinning were analysed separately. Each stage introduced its own acoustic characteristics and potential design challenges. Water movement, pump operation, motor behaviour, valve activity, and the interaction between clothes and the drum all contributed differently to listener perceptions. Individual sound components were isolated and manipulated. Participants evaluated these variations through listening tests, allowing researchers to identify which sounds influenced acceptability most strongly.
The resulting data could then be analysed using statistical models that linked changes in specific sound components to listener ratings. One of the most interesting aspects of this work involved the creation of response-surface models that allowed engineers to visualise how perceived quality changed as different sound characteristics were adjusted. Rather than producing a simple pass-or-fail result, the models created maps of possible design outcomes. Engineers could explore how increasing one characteristic while reducing another might influence listener responses. Product sound quality rarely involves finding a single perfect solution. Designers must balance acoustic quality against manufacturing constraints, performance requirements, reliability considerations, and cost limitations. Statistical modelling provides a way of navigating these trade-offs while retaining a clear understanding of how design decisions influence perception.
Similar principles appeared in Bowen’s work on vacuum cleaners. Consumers often claim that they want quieter products, yet a vacuum cleaner that becomes almost silent introduces a different problem. Users may begin to question whether it is working properly. Certain sounds communicate power, airflow, and cleaning effectiveness. Eliminating every sound is not necessarily desirable. In this case, the challenge is not simply reducing noise but preserving those aspects of the sound that contribute positively to the user’s perception of performance. What emerges from Bowen’s examples is a view of product sound that differs significantly from traditional approaches to noise control. Sounds are not merely by-products of mechanical systems. They communicate information about performance, condition, reliability, quality, and identity. A washing machine, a vacuum cleaner, a refrigerator, and an aircraft each occupy different places in people’s lives. Listeners bring different expectations to each. A refrigerator should not sound like a lawnmower. Equally, a lawnmower should not sound like a refrigerator. The challenge is therefore not simply reducing sound, but designing sounds that make sense within a particular context.
Seen in this light, product sound quality becomes a remarkably human problem. Engineers can measure sound with extraordinary precision. Researchers can develop increasingly sophisticated psychoacoustic models. Statistical techniques can reveal relationships between acoustic characteristics and listener preferences. Yet none of these tools removes the need to understand people. Sound quality emerges not from products alone but from the relationship between products and listeners. What emerged from the lecture was a challenge to a familiar engineering instinct. Faced with a difficult problem, engineers naturally seek better measurements. Bowen’s work suggests that measurements remain essential, though they are not enough on their own. Product sound quality exists at the point where physical acoustics meets human perception. This creates an unusual situation. Few areas of engineering depend so heavily upon subjective judgement while simultaneously demanding rigorous measurement. Product sound quality requires microphones, analysers, statistical models, listening tests, psychoacoustic theory, and human listeners. Remove any one of these elements and the picture becomes incomplete.
Perhaps this is why the question that opened the lecture remains so difficult to answer. Can sound quality be measured? David Bowen’s career suggests that the answer is both yes and no. Sounds can be measured with extraordinary precision. Human responses can be studied, modelled, and predicted. Yet quality itself ultimately emerges through experience. The most successful products are not necessarily the quietest products, nor the products with the best acoustic measurements. They are the products whose sounds make sense to the people who use them. In the end, product sound quality is not really about sound at all. It is about understanding listeners.