An in vivo test for ease of breaking of scored tablets was developed. Scored tablets covering a wide range of dimensions, type of break-mark and ease of breaking were used as training set. Test panels of healthy volunteers (25-61 years old), and panels of elderly (mean age > or =75 years old) were used. Five different test procedures were investigated. Subjective assessment of ease of breaking appeared more cumbersome than objective scaling in "breakable" and "not breakable". Elderly were far less able to break the tablets than healthy volunteers. So, healthy volunteer panels are not a good substitute for the "worst case" patients situation. A test procedure is proposed specifying that not less than 80% of a panel of elderly (mean age > or =75 years old and none younger than 65 years old) must be able to break the scored tablet, with a confidence of not less than 90%.