摘要：Mauri (2010), in her response to our paper (Hengeveld and van Lier 2010), states that adding the connectivity constraint to the sets of constraints we propose in our implicational map of parts of speech would increase the accuracy of our model. Mauri's point is well taken. The connectivity constraint has strong empirical support in a wide variety of semantic domains, and indeed, if this constraint is added to the ones we propose, we can exclude two systems that were predicted by our original model but have not been attested. Thus, with the connectivity constraint added, the model would predict 15 possible systems, 13 of which are attested either in their 'pure' form or in combination with another predicted system. This leaves us with two predicted but as yet unattested systems. These remaining systems are given in Figures 19 and 22 in our contribution to this volume, and are repeated below for convenience. The system in Figure 19 is perhaps just as plausible as a system which has verbs only (represented in Figure 14 of our contribution to this volume). In languages of both these types, lexemes are used for predication, while reference is established indirectly, i.e. through the predication of properties and relations in which a referent is typically involved
