All attitude filter designers are familiar with the covariance matrix tuning process for an attitude filter to get an estimate close to the optimal (or sub-optimal) values as much as possible. This is a difficult and time-consuming task. Selected values for the measurement and process noise covariance matrices are decision makers for the filter's performance. Composing the measurement covariance matrix (R) is rather straightforward as we make use of the sensor specifications. Yet when there are issues such as measurement delay, the matrix components must be tuned for higher estimation accuracy and usually with an ad hoc approach. As to the process noise covariance matrix (Q), we have an analytical approximation, but most of the time it does not give the best result in practice. This paper first surveys researches on tuning the attitude filter and discuss the details of the problem. Then investigates the applicability of common adaptive methods for attitude filtering and provides a performance comparison in between intuitively and adaptively tuned attitude filters.