1- Morning session

About letter shapes - Is there any way to deal with statistical letter shapes? Melissa’s work did build statistical models of letter shapes for the Vindolanda ink tablets, based on positional probabilities within a character box. Another statistical modeling method would be by Principal Component Analysis, provided that enough samples are available. One question is then: “How much do letter shapes differ between ink tablets and incised tablets?”, and also “How could one ’compensate’ for the shapes not represented in the statistical model?” ’Compensating’ isn't quite the right term; what we want is in fact to inform and feed the database (and the model) with the newly found instances of letter shapes as well as consult it. For example in the case of the Frisian tablet (shown during the presentations), dated as a 1st Century AD tablet, a quite unique palaeography was found; ‘A’ as shaped in the document, would not have been identified as ‘A’s according to our current knowledge of shapes of ‘A’s in 1st Century AD Old Roman Cursive.