Despite their general similarity of social organisation, the Greeks and the Romans had adopted opposite solutions to the problem of accommodating freedmen in the society into which they had been manumitted. The Greeks excluded them from citizenship, relegating them to a collective state of dependence as resident aliens; but the Romans on the contrary accepted them as citizens and subjected them to individual ties of dependence.

The dependent status of the early Roman freedman arose from his incorporation into the manumitter's household. All members of the early Roman household were subject to the paternal authority of the head of the household, the paterfamilias. The paternal power of the Romans was unique in that it lasted as long as the pater was alive, not simply until his sons come of age, as is normally the case. All his descendants were thus subject to it regardless of their age and sex, as was his wife and other members of the household such as slaves and freedmen. All were thus bound to obey them, and what is more, all were deprived of individual rights in private law. In particular, persons in potestate could not own, but only hold a peculium which automatically reverted to the pater on their death. There was no difference between the freedman and the filiusfamilias in this respect, nor did such public office as either might enjoy affect their status in private law.

Goldziher regarded Roman law as ‘one of the chief sources of Islamite jurisprudence’ (‘Principles’, p. 296). He first stated this in his article published by the Hungarian Academy of Sciences in 1884 (‘Jogtudomány’); he returned to it in his Muhammedanische Studien, published in 1889–90 (cf. vol. I, p. 188n, vol. II, pp. 75f), and he reaffirmed it in his review of Savvas Pacha in Byzantinische Zeitschrift 1893. FitzGerald wrongly lists Savvas as an adherent of the theory of Roman influence on Islamic law. Savvas used to adhere to this theory, as he himself explains (‘L'erreur, en effet, est si facile!’); but his book was based on his new realization that in fact it is exclusively derived from the word of God and the conduct of the Prophet (Savvas Pacha, Études sur la théorie du droit musulman, vol. I, Paris 1892, pp. xviff, xxi). When Goldziher insisted on Roman influence in his review, debiting Savvas' naiveté to his Oriental origins, Savvas wrote a vehement reply, affirming his position on the origins of the Sharī'a and pointing out that whereas he himself [a Greek Christian] was an Aryan, Goldziher [a Hungarian Jew] was a Turanian whose aggressiveness arose from the fact that he still had some drops of Mongol blood in his veins! (Savvas Pacha, Le droit musulman expliqué, Paris 1896, p. 26).

I shall now try to show that there was no institution similar to Islamic walā' in pre-Islamic Arabia; more precisely, I shall try to demonstrate that four diagnostic features of the Islamic institution were unknown to the homeland of the conquerors, these being the following:

1. The Islamic patronate is individual. It binds one person to another in a relationship of dependence. In pre-Islamic Arabia dependent relationships were formed between groups, or between individuals and groups, not between individuals. In short, in pre-Islamic Arabia such relationships were collective.

2. The Islamic patronate detaches the client from his natal group and incorporates him in that of the patron as a passive member. The Berber client is no longer a member of his Berber tribe, but nor is he a full member of an Arab one. It could be argued that it is enslavement and conversion which detach the client from his kin rather than the institution of walā' itself: the contractual client who remains an infidel presumably remains a member of his natal group as well. But it is to the patron that the latter's group that even the free, unconverted client belongs for purposes of life in Muslim society, and his dependant status in this society arises from the fact that his membership of the patron's group is partial: the patron and his agnates acquire rights and duties vis-à-vis him, but he typically acquires none vis-à-vis them. Individual ties of clientage do not always affect group affiliation, and collective ones never do. Where clientage is collective, the client remains a full member of his own inferior group, as opposed to becoming a partial member of a superior one. Collective clientage creates satellite groups, not semi-members.

3. […]