This initial form of the Teacher
Stress Inventory was used to survey both regular and special education
teachers from Vermont. Prior to this study, 12 conceptually similar
items were added to the factors whose reliability estimates proved
lowest, for a total of 42 items. This longer version of the TSI, using
the same two Likert-type scales to assess the strength and frequency
dimensions, was then distributed to the Vermont regular and special
education teachers during the 1980-81 public school year. Based on
the responses gathered from these teachers, an additional round of
factor analyses was conducted. Using these new data, and based on the
same factor and reliability analyses used with Connecticut teacher data,
one of the additional items was deleted from the item pool, resulting in
a 41-item form. Factor patterns and structure and alpha reliability
estimates were nearly identical to those found in the Connecticut study.
Thus, the 41-item TSI form was established (Fimian, 1984b).
Concurrently, and based on the content appraisal feedback obtained from the 226 stress experts described later in this chapter, one additional a priori factor consisting of 8 items related to time and time management problems was added to the TSI. These items, adapted from Truch (1980) and modified to reflect teaching concerns, were termed "Time Management" and added to the TSI in early 1982, increasing the item n from 41 to 49. Since then this longer form has been used with 17 additional samples.
In this initial form of the TSI there were two dimensions being assessed: strength and frequency. Because the two dimensions were moderately to strongly correlated to one another (Fimian & Zacherman, 1987), some TSI users opted to use only one or the other dimension by using specially developed strength-only or frequency-only TSI forms to assess only the one preferred dimension and to make inferences about the other. Were the TSI strength and frequency dimensions adequately related to one another to warrant the use of either the strength-only or frequency-only form in place of the strength-and-frequency version? Fimian and Zacherman (1987) found substantial positive correlations between the frequency and strength subscale and scale scores. Based on data collected from 14 samples, subscale correlations ranged from a low of .30 to a high of .99, with only one nonsignificant correlation out of over 90 possibilities; in only 1% of the time did a frequency factor not correlate significantly with its strength analogue. Similar findings were evident with respect to the relationships between the Total Strength and Total Frequency scores. In light of these data, the frequency dimension was dropped from further use during the Fall of 1987, and final factor and reliability analyses were conducted at the time on the aggregate data pool (n = 3,401) using only the strength data.