Nigel's language was studied continuously, and the description of it was recast every 1: 1/2 months (Figures 1-6). This was the interval that appeared to be optimal : with a longer interval, one might fail to note significant steps in the progression, while, with a shorter one, one would be too much at the mercy of random non-occurrences in the data. Table 1 (p.147) shows the number pf options within each function at each stage from NL1 (9-10 1/2 months) to NL 5 , the end of Phase 1 (5-6 1/2 months). Those for NL 6 , which is considered to be the beginning of Phase 2, are added for comparison, although it should be stressed that they are not only less reliable but also, as will emerge from what follows, less significant as an index of the system