The relationships between sounds and their perceived meaning and connotations are complex, making auditory perception an important factor to consider when designing sonification systems. Listeners often have a mental model of how a data variable should sound during sonification and this model is not considered in most data:sound mappings. This can lead to mappings that are difficult to use and can cause confusion. To investigate this issue, we conducted a magnitude estimation experiment to map how roughness, noise and pitch relate to the perceived magnitude of stress, error and danger. These parameters were chosen due to previous findings which suggest perceptual congruency between these auditory sensations and conceptual variables. Results from this experiment show that polarity and scaling preference are dependent on the data:sound mapping. This work provides polarity and scaling values that may be directly utilised by sonification designers to improve auditory displays in areas such as accessible and mobile computing, process-monitoring and biofeedback.