The role and reliability of the sensory tests in the emission tests for the emission classification of building materials in Finland

Very important issue for all activities for better IAQ in Finland has been the introduction of the classification guidelines for indoor air quality and climate. These guidelines, published by FISIAQ, including measurable target values, cleanliness requirements and emission criteria's for building materials, has been in use in Finland since 1995 and revised year 2001. An essential part of the successful IAQ classification has been the emission classification of construction materials. The number of building and finishing materials accepted and labeled to the best category, emission class M1, has already exceeded 600, providing a wide selection of products covering all the major types of building materials. The scope of this classification has been widened year 2001 to the cleanliness classification ducts and other components of ventilation systems. In the Finnish M1 classification system the requirements for the emissions are basing as much as possible on well-known, if only possible internationally accepted, testing methods. So requirements are set to the total VOC-emissions, TVOC. This TVOC requirements are however complemented with the requirement to identify the most relevant VOC components, 85% of the highest peaks in the spectogram. Limits are set also to the emission of formaldehyde, ammonia and carcinogenic compounds. Besides chemical tests there are also complementary, independant, sensory tests. These sensory tests are carried out by untrained sensory panels using normal acceptability scale. The probable error in this procedure of the sensory tests using small two step (5/15) untrained panels. The risk connected with the use of these small, untrained panels have been calculated using quite simple basic statistics. These risk calculations are basing on the use well-known T-distribution. In the Finnish M1 emission labelling testing the risk of false decision was chosen for to be 10% and it is equal to both parties. The risk that an acceptable product will be rejected is equal to the risk, that an unacceptable product will be accepted. The calculated accuracy of the mean vote of acceptability is about 0.2 in the used acceptability scale from -1 to +1, when the size of the panel is 15 persons and about 0.4 if the size of the panel is only 5 persons. The calculation procedure of the probable error and the influencing factors are described in this paper.