**4. Conclusions**

The objects discussed in this paper are all first-order prototypes and have been designed to support experimental evaluation with neurotypical participants, rather than patients. Patients with neurological deficit, as the result of degeneration, damage or injury, can often struggle with Activities of Daily Living (ADL). These problems can range from confusion over the sequence in which tasks should be performed, to forgetting which action can be performed using a given object, to failing to recognize a given object. These errors could involve putting breakfast cereal into a coffee mug (when making coffee) or attempting to pour from a milk container before opening it [40]. Around 1/3 of stroke survivors have difficulty with ADL and the errors that they make in planning and execution of errors can be predicted [41]. If it is possible to predict errors (rather than these being random occurrences) then it is possible to predict *when* an error could be likely to arise and to provide some cueing to either prevent the erroneous action or to encourage a preferred, i.e., non-erroneous, action.

Returning to the basic specification for this paper, for a system to provide coaching for patients with neurological deficit to allow them to perform ADL, we would need to recognize which actions people are performing (which typically means having some means of sensing that an action is being performed), to determine the quality of the performance (which requires not only activity recognition but also some evaluation of how this performance meets some criterion), to evaluate performance in terms of an anticipated outcome or intention, and to provide some means of guiding, cueing or otherwise providing advice and feedback in response to the action and in anticipation of subsequent action.

**Author Contributions:** Baber wrote the paper and designed the experiments; Khattab (with Baber) designed and built the objects, and conducted the experiments; Khattab and Baber conducted the analysis of the results; Russell, Hermsdorfer and Wing managed on the CogWatch project that contributed to the design concepts and commented on drafts of the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.
