With the rise of large herbivore populations in most northern hemisphere forests, browsing is becoming an increasingly important driver of forest regeneration dynamics. Among other processes affecting the regeneration, the concept of plant-herbivore feedback loops holds that browsed saplings are more subject to subsequent herbivory. This phenomenon is interpreted as a consequence of compensatory growth following browsing since fast growth is generally associated with higher digestibility and lower defense against herbivores. However, studies linking browsing-induced trait variations to subsequent attractiveness to herbivores are still lacking, especially in the forest context. In this study, we experimentally examine the existence of a feedback loop between oak (Quercus robur L.) and roe deer (Capreolus capreolus) and investigate its underlying morphological and chemical traits. We simulated single and repeated roe deer browsing on nursery-grown oak saplings and measured the changes in sapling height growth, lateral branching, leaf traits and winter shoot traits over two years. We conducted winter feeding trials with tame roe deer one year after the first treatment to test the effect of simulated browsing on sapling attractiveness. Simulated browsing reduced sapling height growth but had no effect on branching. Simulated browsing had no effect on leaf traits after half a year, but decreased the phenolic content and increased the fiber content of winter shoots the following winter. Contrary to our predictions, roe deer preferentially browsed control saplings over saplings previously browsed. After two years, repeated browsing promoted fast carbon acquisition leaf traits (high chlorophyll, high specific leaf area and low fiber content), reduced leaf phenolic content and increased leaf digestibility. We showed that a reduction in 1-year-old oak sapling height growth following browsing, combined with increased structural defense at the expense of chemical defense in winter shoots the following winter, was correlated with reduced browsing pressure, thereby challenging the feedback loop hypothesis. However, we also demonstrated that repeated browsing promoted fast carbon acquisition leaf traits in 2.5-year-old saplings, which tend to support the existence of a feedback loop on older and more intensively browsed saplings. As such, our study provides empirical evidence that morphological and physiological trait responses to browsing influence oak sapling attractiveness, but that the direction and magnitude of this effect depend on the ontogenic stage of the sapling and on the number of browsing events.