8. Conclusions

1. For each tested pattern (m-m, m-n, m-T, N-m, n-n, n-T, T-n), there are far more paradigmatic languages than non-paradigmatic ones (contra WALS for N-m). This suggests that personal pronouns tend to keep their general shape throughout all paradigms.
   
   
2. Given the relative frequency of m, N and T form-types, no correlation could be found for any particular pattern.
   
   
3. Both “WALS patterns” are less frequent than expected in Africa. We could think of two logical explanations:
      - a structural constraint (either inherent or resulting from common diachronic processes), which seems unlikely.
      - the overrepresentation of one or more other patterns that exclude them. So far such patterns were not found. Possible candidates are patterns where one or both of the elements show no root consonant (for ex. *U for 2sg, a major feature of Benue-Congo).
   
   
4. The frequent presence of the m-m pattern calls for testing outside Africa, in order to check whether it is specifically African or not.
   
   
5. But on statistical grounds it can be argued that m-m lgs are not particularly frequent. On the other hand they seem to be typical of the Niger-Congo stock.
   
   
6. > A more refined statistical work is in order.