Details of the studies reviewed

People may be interested in the percentages of filtration. It is important to realise that few of the experiments used the standard methodologies that are used to certify medical masks or respirators (N95s).  Different studies use different methods, different particle sizes, and different ways of counting particles.

Many of these studies were done to provide proof-of-concept rather than to directly inform the design of a mask for wear in the community. 

Different methods

  • There are experiments on flat cloth. Tests using flat material are, actually, the methods used to certify medical masks according to standards.  No additional tests required for the material made into masks.

  • There are experiments, using cloth masks, on outward protection, also called protection of the environment, or source control.  These usually involve volunteers wearing the mask, but some involve a manikin or human-sized bust which generates particles inside the mask.  These methods compare particles inside and outside the mask, or with the mask and without the mask.

  • There are experiments, using cloth masks, on inward protection, or protection of the wearer. Like studies of outward protection, these are done with volunteers or with a manikin.

In all of these types of experiments, the flow rate across the cloth differs from study to study making head-to-head comparisons difficult.

Different particle sizes

The virus that causes COVID-19 is between 0.065 and 0.140 µm in diameter. A micrometre (µm) is one thousandth of a millimetre. In respiratory secretions, the virus is found in particles, which have a range of sizes.  For coughing and sneezing, the most common particle is above 1 µm and some sources suggest may be 10 or 100 µm.  Each experimental set up defines the particle size used for testing.

Different ways of counting particles

In some studies, the actual number of particles is counted.  In other studies, there is a biological marker in the particles (usually a bacterium, sometimes a virus), and what is counted is the number of bacteria or viruses that are transmitted.

All this means that it is difficult to say what percentage cloth filters without giving more information about the study.  And the percentages that we hear people talking about may not be measured in the same way, so that one statistic may not be comparable with another.

Here are some highlights that we found, giving the details of how they were tested. 

In our article, we have noted all the results, whether the cloth provided good filtration or not.  We think what is most useful to people choosing and making masks is to pull out the details of the experiments where cloth provided useful filtration

Flat cloth

For flat T shirt material tested with an aerosol of unknown size, there was 69% filtration of a bacterial marker with one layer, 71% filtration of a bacterial marker with two layers, and 51% filtration of a virus marker with one layer.  (Davies 2013)

For flat linen tea towel (a plain woven cloth), tested with an aerosol of unknown size, there was 83% filtration of a bacterial marker with one layer, 97% filtration of a bacterial marker with two layers, and 72% filtration of a virus marker with one layer. (Davies 2013)

For cotton 600 threads per inch, weight unknown, tested at an unknown flow rate with a particle size of 0.075 to 0.100 µm, there was 76% filtration of a saline particle with one layer, and 85% filtration of a saline particle with two layers. (Konda 2020)

For a hybrid, one layer of cotton 600 threads per inch, weight unknown, and a second layer of 65% flannel, 35% polyester flannel, 90 threads per inch, measured at an unknown flow rate with a particle size of 0.075 to 0.100 µm, 95% filtration of a saline particle with the two layers together. (Konda 2020)

Outward protection, or source control

For a 2-layer T-shirt mask with a sewn edge (prevents stretching and the material opening up), measured with counting transmission of mouth bacteria, categorized by particle size, from volunteers coughing, filtration of 79% overall, and filtration of 71% for particles in the size range 0.65 – 1.1 µm.  In this experiment, a medical mask was tested in the same way and the filtration was 85% overall and 86% for particles in the size range 0.65 – 1.1 µm.  There was no statistically-significant difference between these numbers, suggesting that this cloth mask may be performing similarly to the medical mask tested. (Davies 2013)

A mask made of outer layers of muslin, a plain-woven unfinished cotton cloth, with an inner layer of flannel, measured with counting transmission of mouth bacteria, reduced contamination of surfaces by > 99%, total airborne bacteria by > 99%, and airborne bacteria in particles less than 4 µm (i.e., in the range sometimes called ‘aerosols’) by 88-99%. (Greene and Vesley 1962)

A duckbill-shaped mask made of 4 layers of muslin, measured with counting transmission of mouth bacteria, reduced airborne bacteria overall by 99%, and bacteria in particles less than 3.3 µm by 89%.  This was comparable to the results for commercial disposable masks of the day (1970s) tested in the same way, which showed overall reductions of 96-99%, and for particles less than 3.3 µm, 72-89%. (Quesnel 1975)

Inward protection

A two-layer T-shirt mask with a sewn edge (prevents stretching and the material opening up), measured with saline particles of 0.02 – 1 µm, provided inward filtration efficiency of 50% during a range of activities.  This was lower than observed with a medical mask, which produced filtration of 80-86% when studied in the same way. (Davies 2013)

A one-layer tea towel mask, measured with saline particles of 0.02 – 1 µm, provided inward filtration efficiency of 55 - 77% during a range of activities. (Van der Sande 2008)

Animal experiments

Rabbits wearing 3 or 6 layers of tightly fitting cotton masks, 84 threads per inch, exposed to air particles containing tubercle bacteria (size not given but described as ‘aerosol’), developed fewer tuberculosis lesions than those who did not wear masks. The reduction in tuberculosis lesions was 95%.  This result was statistically significant. (Lurie & Abramson 1949)