alarms resulted in increased response rates to the central alarm. From the graph it appears that, in fact, there may also be an interaction, where there are even larger differences during occasions of lower cognitive relatedness rather than higher. However, a significant interaction was not found. Results indicate that alarm number may overcome the confirming strategy used during situations of low functional component dependency. Although people knew the components were functionally independent, an increase in overall alarm number resulted in a deviation in this cognitive strategy. No other effects were found for responses made.
From viewing Figure 2, it appears that response confidence steadily increased as the assumed functional dependency between component pairs increased. A trend analysis revealed that there was, in fact, a positive linear relationship between functional dependency and response confidence F(1,22)<14.78, p<.001. Moreover, a four-factor mixed ANOVA revealed reveal a significant effect for experience on response confidence F(1,20)=5.27,p<.05. As participants gained experience responding to the alarm, they became more confident in their adopted strategy. Response confidence increased in spite of no performance feedback being provided to participants. Also, even though no differences in response strategies were found between automobile experts and novices, a difference between these groups existed in terms of their response confidence, where those who were more knowledgeable in mechanics tended to be more confident in their responses F(1,20)=5.70, p<.03.
It was found that participants' response rates to a 60 percent true alarm increased as the degree of functional dependency between two implicated components increased. Response rates ranged from well below to well above 60 percent in a linear fashion as the degree of agreement between components increased, demonstrating a systematic heuristic based upon cognitive relationships for all who participated regardless of expertise in automobile mechanics. This means that although participants may have ranked the same component pairs as having a different level of dependency, they all tended to use the degree of dependency