“…On the other hand, multiple relaxations of parameter constraints at Step 1 that are guided primarily by fit indexes (e.g., modification indexes) are not recommended. A number of studies have shown that repeated respecifications of this sort will fail to lead to the "true" model in many cases, and that the chosen modifications can vary widely across samples (Kaplan, 1989(Kaplan, , 1990MacCallum, 1986;MacCallum, Roznowski, & Necrowitz, 1992). Our view is that if modifications to the Step-1 model are introduced in this manner, the resulting modified model should be regarded as an exploratory creation that will need confirmation in future samples.…”