This is not a very strict agreement procedure, since a total of 100% could be determined if two observers recorded totally different cases of target responses within the same 15-metre observation. In the data flow example shown in Figure 1, Observer 1 records three target response instances during the 3 m (one per minute) of its observation, two instances per minute 4 and misses all other instances for the remaining 12 meters. During the same hypothetical observation, Observer 2 missed all three instances during 1-3 minutes, recorded a target response instance in minute 4, but recorded four instances in minute 15. Although these are totally different events, the total number of IAOs that would result would still be 100%. Figure 2 presents the results of each observer for each of the six types of meetings, and Table 1 summarizes the results in average percentages for observers. All computational methods showed high levels of reliability in moderate and high sessions, suggesting that higher response rates did not affect reliability per se. Similarly, all computational methods provided high reliability values in constant and burst meetings, suggesting that the response did not appear to have an impact on reliability, at least for these observers. Reliability values for the mid-interval session were also uniformly high, based on all calculations. The only significant discrepancy between the reliability indices was observed during the half session. Overall reliability remained high (M – 99.3%), while lower values were gradually established by intervals, proportionalities and exact methods (Ms – 86.3%, 71.9% and 53.7%, respectively). This effect was extremely consistent for all observers, but it was most pronounced for observers 6, 7 and 10. An alternative to overall reliability is the reliability of the interval, which consists of dividing a session into small time intervals of the same length and comparing observer records during each interval (Bijou et al., 1968).
An agreement is defined as both an observer who records at least one answer or no response at a given interval, and reliability is calculated by determining the intervals of the number of chords by the number of intervals in the meeting.