Studying the microclimate of plant canopies has long motivated scientists in various research fields such as agronomy, ecology or silviculture, and almost a century has passed since the first measurements of wind speed in a forest stand were published in the scientific literature. The behaviour of wind in canopies is an essential component of their microclimate, which largely conditions the rate of exchange of heat, water vapour, and other gases and particles of interest with the atmosphere. This review examines the evolution of our understanding of turbulent flow in plant canopies, focussing on the period that covers the last fifty years (1970-2020). We first describe how our knowledge and ideas have evolved since canopy flow became a topic of interest, and show how the 1970s was a pivotal decade in this field. Until then, canopy turbulence was considered to result from the superposition of standard surface-layer turbulence and small-scale turbulence generated in the wakes of plant elements. However, it was progressively found that the flow in plant canopies is dominated by large coherent structures, giving canopy turbulence unique characteristics. We thus describe the particular nature and structure of canopy flows, based on experimental observations accumulated over several decades. We show how canopy turbulence was reconsidered on the basis of a now widely-accepted analogy with a plane mixing layer, and we examine the significance of a key parameter, the "canopy-shear length scale". Investigating the effects of canopy density and atmospheric stability, we then discuss the extent of the mixing-layer analogy and the limits of our current understanding of canopy turbulence. Finally, we review the modelling tools used in this field and show how their development has evolved to date to meet our needs. In conclusion, we present a historical summary of the evolution of this research field and suggest future directions.