Samuelson introduced the term “neoclassical synthesis” in 1952 (1952: 60). He used it later in the third edition of his Economics to refer to the integration of national income analysis and traditional price theory (Backhouse 2014). From the sixth edition, when growth theory became a major issue for most economists in the 1960s, the synthesis took on a specific meaning and began to be understood as describing the outcome of a process in which it was claimed short-run Keynesian and long-run neoclassical analyses were compatible. In a nutshell, as long as it was assumed that the economy was managed on a Keynesian-basis in the short-run, the neoclassical growth model developed by Solow (1956) and Swan (1956) was seen as the most appropriate tool to analyze and sustain full-employment growth. This notion was quickly challenged through debates on the dynamics of income distribution and expectations and finally also questioned by Arrow (1967). Following Arrow’s intuition and using original archive material from Duke and Cambridge Universities, this paper proposes to clarify the political and theoretical dimensions of Samuelson’s synthesis and the debates that ensued.