Conceptually and methodologically, the minimal effects models have limitations. Much of the empirical data was gleaned from research on the media’s impact on voters during elections in the US. Using voting as the dependent variable poses problems. Chiefly, it does not measure effect accurately. While the media may not change a voter’s decision, it can still influence the voter’s support. If, after consuming the media messages, a voter is more convinced than ever, that in itself is still an effect. Sometimes, a voter’s confidence in a candidate may be weakened but not to the point of voting for the candidate’s rival and this effect (of weakened conviction) will not show up in the data. The primary method of data collection for these studies are surveys, and this has been criticised as unreliable, as the voters are required to recall and report their own vote as well as decision making process. Thus, they have to rely on their memory of what are likely to be transient moments. For example, they are unlikely to be aware of what they were thinking of when listening to a political debate or advertisement.