Humans continuously adjust their actions to deal with a dynamic environment. For example, we correct for motor errors or perceptual misjudgements in a process called sensorimotor adaptation, which can occur both with the actor detecting (explicitly) and not detecting it (implicitly). Experimentally, sensorimotor adaptation is often studied by creating mismatches between sensory channels by applying perturbations to one sensory modality. In our experiment, we investigated how the size of the perturbation and the error signal (dissociated by using either a sinusoidal perturbation function or a step function) each contribute to the detection of the perturbation by the participant. Specifically, participants were asked to grasp cuboids of different lengths in a mirror-setup allowing us to present different sizes for seen and felt cuboids, respectively. We used a 2AFC task to assess when participants noticed these mismatches and applied a linear state-space model to the maximum grip aperture, modelling error-correction and its retention. We found larger just-noticeable differences (JNDs) when perturbations follow a sinusoidal compared to a step function, consistent with the idea that reduced error signals following adaptation make it harder to detect perturbations. However, no corresponding difference was found in the error-correction parameter, and a weak correlation with the JNDs. Dissociating the visuo-haptic mismatch and the error signal of perturbations using sensorimotor adaptation provides insights into what makes participants aware of sensory perturbations. Knowing when participants know that an action has been perturbed will allow us to get a better understanding of when error correction may be under cognitive control.