The current study undertakes refixation patterns on words in sentential reading. Utilizing a Linked Linear Mixed Model approach, the analysis focused on words with a single fixation and the first fixation from words with a double fixation. The model findings revealed a relationship between refixation probability and fixation locations, with initial fixations tending to occur closer to the beginning of a word in instances of higher refixation likelihood. Incorporating predicted and residual values of the fixation location models into the fixation duration models resulted in congruence in the observed fixation locations, durations, and residual values. Finally, the models revealed differences between progressive and regressive second fixations.