Task difficulty (TD) reflects students' subjective judgement on the complexity of a task. We examine the TDs data of 236 undergraduate students in a simulation-based "Predict-Observe-Explain (POE)" environment using three different labels "easy," "medium" and "hard." Generally, the students who perceive the tasks to be "easy" or "hard" perform poorly at the transfer task than the students who perceive the tasks to be "medium" or moderately difficult. Sequences of students' TDs are analysed which consist of a set of several judgements, collected once for each task in a "POE" sequence. The analysis suggests that given a sequence of TDs, difficulty level "hard" followed by a "hard" may lead to poorer learning outcomes at the transfer task. By contrast, difficulty level "medium" followed by a "medium" may lead to better learning outcomes at the transfer task. In terms of the TD models, we identify student behaviours that can be reflective of their perceived difficulties. Generally, the students who report that the tasks are "easy," adopt a trial-and-error behaviour where they spend lesser time and make more attempts on tasks. By comparison, the students who complete the tasks in a longer time by making more attempts are likely to report that the following task is "hard." For the students who report "medium" TDs, mostly these students seem to reflect on tasks where they spend a long time and require fewer attempts for task completions. Additionally, these students provide longer texts for explaining their hypothesis reasoning. Understanding how student behaviours and TDs manifest over time and how they impact students' learning outcomes is useful, especially when designing for real-time educational interventions, where the difficulty of the tasks could be optimised for students. It can also help in designing and sequencing the tasks for the development of effective teaching strategies that can maximise students' learning.