49 Abstract This paper discusses the application of existing methods for measuring reliability to collection condition survey data. The essential requirement of reliability is understood by conservators but has never been quantified. The relatively recent introduction of computer methods that calculate reliability is mentioned as a significant step towards introducing the practice of ensuring reliability during pilot studies. The paper discusses different methods of quantification, and their qualities, as well as the aims of assessing reliability for condition surveys. The most satisfactory methods, and choosing an appropriate level of reliability for conservation applications are suggested. However, no single number is offered as a standard. It is recommended that more than one index should be used. The paper also describes techniques to use indices to determine causes of disagreement, so reliability can be increased. Key words: Condition Survey, Reliability, Index, Data, Measurement The Conservator volume 30 2007 Joel Taylor and David Watkinson Indexing Reliability for Condition Survey Data Introduction Collection condition surveys involve evaluating the condition of large numbers of museum objects that can vary in a number of ways. These assessments are more complicated than they might seem, as many different factors contribute to an object’s condition. The term ‘Condition’, like ‘Health’, can be used in a variety of different ways, which vary according to context. Condition can incorporate a wide range of issues, including stability, need of treatment, utility, completeness, evidence of past damage and vulnerability. These evaluations usually result in the objects being assigned an overall score or grade on a scale of condition. However, the way in which people interpret definitions and information can vary, leading to different approaches to object assessment, even with guidance. As a result, differences of opinion can arise, and it has been demonstrated that one cannot assume condition data to be reliable (Taylor & Stevenson, 1999). A reliability index is a method for giving an indication of the reliability of the data. It uses a formula to provide numerical value to express the agreement between different assessments of the same phenomena. Measuring reliability is a processoftencarriedoutinfieldsthatusesemi-quantitativedata.Thetechniques are used in fields, such as content analysis, marketing and social science surveys (Neuendorf, 2002). They are used to check and improve data reliability of a variety of phenomena. “A reliable procedure should yield the same results from the same set of phenomena regardless of the circumstances of application” (Krippendorff, 1980, p.129), such as surveyors involved, the survey form, the working conditions, the sample chosen and the number of surveyors. Reliability is the extent to which independent surveyors evaluate a characteristic of an object or population, such as condition, and reach the same conclusions. It is a function of the precision of the category definitions, the ambiguity of objects being assessed, and the surveyor’s judgment. How these factors can be analysed separately is discussed later. It is important to recognise that reliability is not a measure of the validity of judgements, although its influence on this is strong, as it determines how much people agree. Existing conservation practice does not currently include measurement of reliability, so these differences of opinion can go undetected. If significant decisions on collection preservation are undermined by the absence of reliability, the extent to which survey data can meet the needs of the conservator needs to be understood and verified. The first step in determining reliability is its expression. Methods to measure reliability, such as the indices that will be discussed here, are not well known and are difficult to carry out by hand. Relatively recent developments have allowed these tasks to be carried out by computer and equip the conservator with a way to assess data reliability. This paper aims to introduce some methods and key issues that surround them, as well as refer to computer software that can make reliability measurement suitable for the practising conservator. The opportunity to measure and improve reliability of condition data has benefits for any institution that carries out such activities. These methods can now be applied to condition surveying practice. There is seldom perfect agreement in semi-quantitative data collection but equipping the conservator with the tools to measure reliability means that condition surveying data can be can influenced in a number of ways. Reliability