A systematic survey of rare-earth (r.e.) abundances in submarine tholeiitic basalts along mid-oceanic ridges has been made by neutron activation analysis. The r.e. fractionation patterns are remarkably uniform along each mid-oceanic ridge and from one ridge to another (Juan de Fuca Ridge, East Pacific and Chile Rise, Pacific-Antarctic, Mid-Indian and Carlsberg Ridge, Gulf of Aden, Red Sea Trough and Reykjanes Ridge). The patterns are all depleted in light r.e. except for three samples (Gulf of Aden and Mid-Indian Ridge) which are unfractionated relative to chondrites. They contrast markedly with tholeiitic plateau basalt which are shown to be related to the early volcanic phases associated with continental drift. Tholeiitic plateau basalts are light r.e. enriched as are most continental rocks. Mid-ocean ridge basalts are also distinguishable from spatially related oceanic shield volcanoes of tholeiitic composition (Red Sea Trough-Jebel Teir Is., East Pacific Rise-Culpepper Island). Thus on a r.e. basis there are tholeiites within tholeiites. The r.e. difference between mid-ocean ridge tholeiites and tholeiitic plateau basalts can be related to distinct thermal and tectonic regimes and consequently magmatic modes and rates of intrusions from the low velocity layer in the upper mantle. The difference between continental and oceanic volcanism appears to be triggered by: (1) presence or absence of a moving continental lithosphere over the low velocity layer, and (2) whether or not major rifts tap the low velocity layer through the lithosphere. Fractional crystallization during ascent of melts before eruption at the ridge crest does not affect appreciably the relative r.e. patterns. R.e. in mid-ocean ridge basalts appear to intrinsically reflect their distribution in the upper mantle source, i.e. the low velocity layer. Based on secondary order r.e. variation of mid-ocean ridge basalts: (1) If fractional crystallization is invoked for the small r.e. variations, up to approximately 50% extraction of olivine and Ca-poor orthopyroxene in various combinations can be tolerated. However, only limited amount of plagioclase or Ca-rich clinopyroxene can be extracted, the former because of its effect on the abundance of Eu abundance and the latter because of its effect on the [La/Sm]$_{\text{e}.\text{f}.}$ ratio, alternatively. (2) If partial melting during ascent is invoked, and a minimum of 10% melting is assumed, the permissible degree of melting of originally a lherzolite upper mantle may vary between 10 and 30%. It is not possible to establish readily to what extent these two processes have been operative as they cannot be distinguished on the basis of r.e. data only. However, there is evidence indicating that both have been operative and are responsible for the small r.e. variations observed in mid-ocean ridge basalts. An attempt to correlate second order r.e. variations along or across mid-oceanic ridges with spreading rate, age, or distance from ridge crests has been made but the results are inconclusive. No r.e. secular variation of the oceanic crust is apparent. R.e. average ridge to ridge variations are attributed to small lateral inhomogeneities of the source of basalts in the low velocity layer, and to a certain extent, to its past history. The remarkable r.e. uniformity of mid-oceanic ridge tholeiites requires a unique and simple volcanic process to be operative. It calls for upward migration of melt or slush from a relatively homogeneous source in the mantle-the low velocity layer, followed by further partial melting during ascent. The model, although consistent with geophysics, may have to be reconciled with some evidence from experimental petrology. Models for r.e. composition of the upper mantle source of ridge basalt, formation of layers 2 and 3, and the moho-discontinuity, are also presented.