Early developments of growth theory are seen widely as the result of a two-step process – the first represented by Harrod's Essay in Dynaamic Theory, and the second by Solow's 1956 model. Harrod is considered to be the first to highlight the pervasive instability in macrodynamics, which Solow showed disappeared with the inclusion of flexible-coefficient production functions. It has been recognized since that this is a misreading (Besomi 1995, 1998, Bruno and Dal-Pont Legrand 2014). Hoover and Halsmayer (2016) examined how this "culture of misunderstanding" guided both Solow's modelling work and his reading of Harrod. Our paper pays attention to the specific issue of the introduction of an (independent) investment function in those early growth models. Using new archival material, we examine this complex issue and show how macroeconomists of that period dealt with problems related to incorporating expectations, an a priori unavoidable step in order to build robust investment functions. Those elements were indeed discussed at length, in the early 1960s, by economists such as Sen, Samuelson and Solow as shown in his correspondence with Hahn. Our paper sheds light on some hidden foundations of growth models and examines the nature of the break Solow’s model introduced in the growth research program as initially defined by Harrod.