Active learning has become one of the most influential ideas in university science teaching. Instead of spending an entire class listening to an instructor, students may work through problems together, test predictions, discuss evidence or complete structured tutorials. Yet a practical question follows whenever a teaching method moves from its developers into ordinary classrooms: how closely must instructors reproduce the original method for it to work?
New peer-reviewed research suggests that the answer may be more flexible than a strict fidelity model implies. A study published in Physical Review Physics Education Research on 2 October 2026 examined how 18 introductory physics instructors implemented three established active-learning approaches. The researchers found that instructors could organise the same critical components in substantially different ways, while their sample showed no clear relationship between observed implementation fidelity and students’ conceptual learning gains.
Why implementation fidelity matters
Evidence that active learning can improve outcomes does not mean every classroom labelled “active” is doing the same thing. Teaching innovations are often developed under particular conditions, with specific classroom layouts, resources, student populations and institutional support. When another instructor adopts the method, some features may be retained while others are modified to fit local circumstances.
This creates a problem for both research and practice. If a modified version produces different results, it can be difficult to know whether the teaching method itself is responsible or whether the adaptation changed something essential. Conversely, insisting that every instructor reproduce a method exactly could discourage useful adaptations that make the approach workable in a different institution or classroom.
Ibukunoluwa Bukola and colleagues therefore focused on the critical components of three named approaches used in introductory physics: Student-Centered Active Learning Environment with Upside-Down Pedagogies, commonly known as SCALE-UP; the Investigative Science Learning Environment, or ISLE; and Tutorials. A critical component was defined as a feature without which the method could no longer reasonably be considered an implementation of that method.
The researchers observed 18 instructors
The team first identified the critical components associated with the three teaching approaches. It then analysed classroom observations from 18 different introductory physics instructors who used these methods and compared their broader implementations with high-fidelity examples.
Rather than relying only on whether an instructor said that they used a particular method, the researchers examined what instructors and students actually did during class. This allowed implementation fidelity to be studied through observable classroom behaviour, including how much class time was devoted to the activities regarded as central to each approach and how those activities were sequenced.
The study also considered conceptual learning gains. This is important because fidelity is ultimately most useful as an educational concept if differences in implementation help explain differences in student learning. A method could look different from its original design while still preserving the mechanisms that make it effective.
Similar time on critical components, different classroom sequences
Across SCALE-UP, ISLE and Tutorials, the broader implementations and the high-fidelity implementations spent similar amounts of classroom time on the critical components identified by the researchers. That finding suggests that instructors who adapted the methods were not necessarily abandoning their defining elements.
Where the classrooms differed more strongly was in how those elements were put into practice. The researchers observed substantial variation in the sequences of activities used to operationalise the same critical components. One instructor, for example, might organise a small number of relatively long activities, while another could use many shorter activities.
This distinction is important. Two classrooms can devote comparable time to the same underlying instructional purpose without looking identical from minute to minute. Fidelity therefore may not be well represented by a simple question of whether a classroom exactly resembles the version originally designed by a teaching-method developer.
No clear fidelity link to conceptual learning gains
The study’s most consequential result concerned student learning. Within this sample of instructors, the researchers found no clear relationship between fidelity of implementation, measured through student and instructor behaviours during class, and students’ conceptual learning gains.
That result should not be interpreted as evidence that implementation does not matter. The study involved 18 instructors, and an absence of a clear relationship in this sample is not proof that every modification is equally effective. Instead, the result provides preliminary evidence that there may be more than one productive way to implement the critical components of an established active-learning method.
The distinction between critical components and surface-level classroom routines is especially useful here. If a method’s educational mechanism depends on students reasoning collaboratively, testing ideas or receiving structured guidance, preserving that mechanism may matter more than reproducing the exact duration and ordering of every classroom activity.
What this could mean for university teaching
For instructors, the findings support a more nuanced view of adaptation. Universities differ in class size, room design, timetabling, staffing, equipment and student preparation. A teaching method developed in one setting may therefore need to change when it is adopted elsewhere.
The study suggests that evaluation should pay attention to whether the defining components remain present rather than treating every departure from an original lesson sequence as a failure of fidelity. This could give instructors room to adapt evidence-based teaching while still maintaining the features that distinguish a named method from generic classroom activity.
For researchers, the findings also highlight why labels alone can be misleading. Two instructors may both report using SCALE-UP, ISLE or Tutorials while organising classroom activity differently. Direct observation can reveal this variation and make comparisons between implementations more informative.
Important limitations
The findings are preliminary rather than a licence for unrestricted modification. The sample contained 18 introductory physics instructors and focused on three named active-learning methods. Results therefore should not automatically be generalised to every university discipline, teaching innovation or educational setting.
The study also examined fidelity through observable student and instructor behaviours. Implementation has other dimensions, including the quality of facilitation, the difficulty of tasks, student preparation and institutional conditions, which may influence learning even when the visible structure of classroom activity appears similar.
Finally, finding no clear relationship between fidelity and conceptual learning gains does not establish that fidelity has no effect. With a modest instructor sample, subtle relationships may be difficult to detect, and some adaptations could still weaken or strengthen outcomes. Larger studies across more institutions and disciplines would help establish where productive flexibility ends and consequential alteration begins.
The broader lesson is therefore not that fidelity is irrelevant, but that fidelity may need to be defined around what a teaching method is trying to make students do rather than around an exact classroom script.
Source Information
Study: Measuring fidelity of implementation of named active learning methods in physics
Authors: Ibukunoluwa Bukola, Meagan Sundstrom, Justin Gambrell, Colin Green, Adrienne L. Traxler and Eric Brewe
Journal: Physical Review Physics Education Research, Volume 22, 020142
Published: 2 October 2026
DOI: 10.1103/xcdt-txzh
The article is open access under a Creative Commons Attribution 4.0 International licence.
