Several tests that can be used to determine dust explosibility parameters have been presented in the technical and scientific literature to assess the characteristics of a dust sample. Some parameters are countable variables that can be defined by a numerical value (e.g. Pmax, kst), while others define a threshold above (or below) which an explosion should not occur (MIE, LOC, autoignition temperature). The latter parameters should be regarded as limit values at which the probability of occurrence of ignition (or explosion propagation, when LOC or LEL are considered) is low enough to be considered negligible. Technical standards describe the procedures that should be adopted to measure different parameters. The Cloud Autoignition Temperature (AIT) is defined, according to the EN 50281 European standard, as the temperature at which 10 consecutive ignition attempts are negative (no ignition). A similar assumption is made in UNI 13821, where MIE is considered. However, these technical standards do not state what the confidence of the threshold value, as determined by standard procedures, is. The ignition probability at such a threshold is assumed to be zero, but this is considered just an assumption, since the relationship between the probability of occurrence of a studied phenomenon (ignition for example) and the independent variable (the temperature, in the case of AIT) has not yet been established. It is known however that the ignition probability does not change abruptly from 0 to 1 as a step function, but through a smooth slope over a temperature interval (Bibbona et al., 2016). The aim of this work has been to propose a new definition of the autoignition temperature that is based on the ignition probability concept. AIT is defined as the temperature at which the ignition probability is lower than a given threshold. In order to arrive at this definition, the dependence of the cloud Autoignition probability on the temperature in the temperature interval close to AIT needs to be investigated. In the present experiment, Lycopodium and Wood Char were used to carry out the tests. The Ignition probability vs temperature relationship in the range in which the probability varies from zero to one (we have called such a range probability escalation interval, PEI) has been determined and the resulting experimental data have been interpreted by means of an appropriate statistical model. Experimental data can, in fact, be used to identify the threshold below which the ignition probability is lower than a suitably predefined level with prescribed confidence, thus providing a more solid ground on which the definition of the AIT could be based.