Abstract

Students in archaeology and the geosciences learn many methods but rarely learn how to combine them into one defensible argument. This paper proposes a five-stage framework that can serve as both a research protocol and a teaching model. The stages are: 1) formulation of a testable hypothesis; 2) organization of a legally authorized, institutionally continuous research program; 3) multidisciplinary investigation across four evidence streams (archaeology and laboratory analyses; geodesy, GIS and remote sensing; geophysics and environmental measurement; spatial statistics and archaeoastronomy); 4) international scientific evaluation; and 5) peer-reviewed publication. A single rule runs through every stage: no finding receives interpretive weight until it has been compared with an explicit benchmark, such as a natural control, a reference standard or a null model. Each stage pairs an educational function with defined learning outcomes, teaching activities, and assessment methods. The paper then outlines applications in university courses, field schools, and early-career research training; a one-semester curriculum with an integrated field module; and a design for future empirical evaluation. The framework draws on about twenty years of multidisciplinary investigation in the Bosnian Valley of the Pyramids. The case is used here as teaching material, including its contested interpretation. The educational model is a proposal. Its effect on learning has not yet been tested and should be evaluated independently.

Keywords

Interdisciplinary Education, Research-Based Learning, Archaeological Methodology, Geospatial Technologies, Scientific Reasoning, Field Schools

Share and Cite:

1. Introduction

Archaeology is now a multidisciplinary science. A single project may involve excavation, radiocarbon dating, petrography, GNSS surveying, LiDAR, ground-penetrating radar, and spatial statistics (Renfrew & Bahn, 2020; Conolly & Lake, 2006). Each method answers a narrow question. None answers the larger one alone. Excavation samples a small part of a site. Geophysics detects anomalies but cannot say what they are. GIS measures spatial relationships but does not explain them. Statistics can show that a pattern is unlikely to be random yet cannot show who made it.

The hard part is integration. How should results from disciplines with different assumptions and standards of proof be combined into one argument? Kintigh et al. (2014) listed integrating heterogeneous data across disciplines as one of archaeology’s grand challenges. The same problem appears in teaching. Curricula usually teach methods one at a time, in separate courses. Students learn to operate a total station or run a Monte Carlo simulation. They are seldom taught how to decide what a result means when it sits beside four other results from four other fields.

Research on higher education points in the same direction. Students gain more from authentic inquiry than from hearing about its results (Healey & Jenkins, 2009; Linn et al., 2015). Undergraduate research experiences are associated with gains in confidence, persistence and research skills (Lopatto, 2004; Russell et al., 2007). Interdisciplinary learning, however, does not happen on its own. It needs explicit structure, so that students learn to connect disciplinary insights rather than collect them (Spelt et al., 2009; Repko & Szostak, 2020).

This paper proposes such a structure. It adapts a five-stage research framework, first developed to organize a long-term investigation in Bosnia and Herzegovina, into a model for teaching interdisciplinary scientific reasoning. The framework keeps its research logic: hypothesis, organization, multidisciplinary investigation, evaluation, publication. To each stage, it adds an educational function, learning outcomes, activities, and assessment.

The framework’s central rule is also its main teaching point. Every finding must be compared with an explicit benchmark before it is interpreted. The benchmark may be a natural control, a laboratory reference standard, a comparison dataset or a null model specified in advance. These benchmarks do different jobs. A laboratory reference standard establishes measurement quality: it shows that an instrument or test gives accurate, repeatable values. It does not show where a sample came from. A suitable geological comparator, such as rock from local natural outcrops tested with the same protocol, helps distinguish natural materials from manufactured ones. Students learn to ask which of the two jobs a given comparison does. This rule separates observation from interpretation. It also gives students a concrete habit to practice at every step.

The framework draws on about twenty years of research in the Bosnian Valley of the Pyramids near Visoko. That program is used here as a case study and a source of teaching material. The paper does not ask readers to accept the program’s interpretation of the site, which remains contested (Section 7). The paper offers the educational model as a proposal. No student data are reported, and no claim is made that the model already improves learning. Section 6 sets out how that claim could be tested.

2. Conceptual Foundations

2.1. Research-Based and Inquiry-Based Learning

Healey and Jenkins (2009) distinguish teaching that presents research results from teaching in which students do research themselves. The second form is associated with deeper understanding and stronger research skills. Inquiry-based learning follows a similar logic. Pedaste et al. (2015) reviewed inquiry models and found a common cycle: orientation, conceptualization, investigation, conclusion, and discussion. The five-stage framework maps closely onto that cycle. In the Pedaste et al. model, the discussion phase comprises communication and reflection, and both can take place throughout the inquiry process, not only at its end. The framework keeps this feature. It extends the model in two specific directions: the institutional organization of research (Stage 2) and communication beyond the classroom through external evaluation and formal publication (Stages 4 and 5).

Guidance matters. Unguided discovery is inefficient for novices (Kirschner et al., 2006). Problem-based learning works best when tasks are structured, and expert support is available (Hmelo-Silver, 2004; Prince & Felder, 2006). The framework therefore works as scaffolding. It tells students what must be done at each stage and what standard the work must meet. It does not tell them what answer to reach.

2.2. Situated Learning and Field Practice

Knowledge in a discipline is learned through participation in its practice (Brown et al., 1989; Lave & Wenger, 1991). In archaeology, the field school is the main site of this participation. It is also where students learn the profession’s habits of recording, sampling, and documentation (Perry, 2004; Mytum, 2012). Field schools often concentrate on excavation techniques. The framework adds the steps that come before and after: formulating the question, planning the sampling, testing results against controls, and reporting them to peers.

2.3. Evidential Reasoning and the Nature of Science

Philosophers of archaeology describe archaeological inference as reasoning from several independent, partial lines of evidence (Wylie, 2002; Chapman & Wylie, 2016). Confidence grows when independent methods converge. The same principle appears in other sciences as triangulation (Munafò & Davey Smith, 2018). Teaching this principle explicitly is part of teaching the nature of science: what counts as evidence, how hypotheses are tested, and why claims stay open to revision (Lederman et al., 2002; Popper, 2002).

Chamberlin’s (1890/1965) method of multiple working hypotheses is especially useful for teaching. Students who hold only one hypothesis tend to look for confirming evidence. Students who must state several competing hypotheses, each with its own predictions, learn to design observations that discriminate between them. The framework adopts this method at Stage 1.

2.4. Constructive Alignment

Biggs (1996) argued that learning outcomes, teaching activities, and assessment tasks should align with one another. The competency matrix in Section 5 follows this principle. For each stage, it states what students should be able to do, how they will practice it, and how it will be assessed. Outcomes are written with verbs from the revised Bloom taxonomy (Anderson & Krathwohl, 2001), so that the expected cognitive level is visible.

2.5. What the Framework Adds

Inquiry cycles, research-based learning and constructive alignment are well established. The framework draws on all three. It extends them in three specific ways.

First, it carries communication into institutional organization and publication. The inquiry cycle synthesized by Pedaste et al. (2015) already includes communication and reflection, and places them throughout the inquiry process. The framework does not replace that feature; it extends it. In the inquiry cycle, communication usually stays within the class. In the framework it leads to the institutional and social stages that make research trustworthy: legal authorization and data stewardship (Stage 2), exposure to external critics (Stage 4) and formal peer review and publication (Stage 5). Students practice these stages rather than hear about them.

Second, it makes the benchmark mandatory. Inquiry models ask students to collect data and draw conclusions. They rarely specify what a result must be compared with. The framework requires a justified comparison for every finding: a natural control, a reference standard or a null model fixed in advance. Students must also argue why the chosen comparator is the right one.

Third, it teaches students to count evidence correctly. In multidisciplinary work, several analyses often rest on the same samples or the same measurements. They look like convergence but are not independent. The framework requires students to record the source of each finding and to treat several analyses of one dataset as a single line of evidence. This distinction is central to triangulation (Munafò & Davey Smith, 2018), yet it is seldom taught explicitly.

Constructive alignment supplies the structure of the teaching design (Section 5). It does not supply the content. These three features are the content.

3. The Five-Stage Framework and Its Educational Functions

The framework has five sequential stages. In research, they move a project from question to publication. In teaching, each stage also develops a specific set of competencies. Table 1 summarizes both functions.

Table 1. The five stages: scientific function and educational function.

Note: The five stages are sequential, but findings at Stages 3 - 5 routinely return the project to Stage 1 for a revised hypothesis.

3.1. Stage 1: Hypothesis Formulation

In research, Stage 1 turns an observation into a question that can be tested. A hypothesis must state what evidence would support it and what evidence would count against it. It must remain open to modification or rejection.

The Bosnian program began with the hypothesis that several pyramid-shaped hills near Visoko were not exclusively natural landforms but contained anthropogenic components. The hypothesis concerns large-scale modification, not ordinary human use of a hill. Stated precisely, it holds that 1) the main faces of the hills were deliberately shaped into planar surfaces of regular orientation; 2) parts of these faces and of the terraces are covered with artificially produced or deliberately placed blocks and slabs; and 3) this work extends over areas of thousands of square meters on each hill, not over isolated spots. Evidence of ordinary human activity on a natural hill does not test this hypothesis. Visočica, for example, carries the remains of the medieval royal town of Visoki, and hills in the region also show paths, agricultural terraces, small quarries and burials. Such traces are expected under the natural-landform hypothesis as well. Only observations that separate a shaped and clad hill from a natural hill with ordinary human use count as tests. This hypothesis generated questions about morphology, construction materials, chronology, subsurface structure and spatial organization. Each question could be addressed by a different discipline.

The educational function of Stage 1 is critical thinking. Students practice three skills. First, they convert a broad observation into a precise, falsifiable statement. Second, following Chamberlin (1890/1965), they write competing hypotheses. For a hill of regular shape, these would include a fully natural origin (for example, tectonic uplift and differential erosion), a natural landform with ordinary human use (settlement, fortification, terracing), a natural landform deliberately shaped and clad on a large scale, and a largely constructed feature. Third, they list, for each hypothesis, the observations it predicts and the observations it forbids.

The Visoko case is well suited to this exercise because the competing hypotheses are real and are argued in the literature. Students can be asked to write the predictions of the natural hypothesis as carefully as those of the anthropogenic one. They then identify which planned observation would best separate the two. The exercise shows that a good hypothesis is defined as much by what it rules out as by what it proposes.

3.2. Stage 2: Research Organization

Long-term multidisciplinary research needs legal authorization, stable institutions, standard documentation, and clear responsibility. Short excavation campaigns rarely produce the continuous datasets that complex sites require. In Bosnia, the Zenica-Doboj Canton issued excavation permits between 2006 and 2012. A registered foundation carried out the work, opening 20 trenches between 2005 and 2025 under continuous institutional authorization.

Research organization is seldom taught as part of the scientific method. It should be. A project without permits, sampling plans, archives and data standards cannot produce evidence that others can check. The educational function of Stage 2 is research ethics, institutional responsibility and project management.

Students learn how heritage law governs excavation and sampling. They write a short research design with a sampling strategy, a documentation protocol and a data management plan. Data management should follow the FAIR principles, so that data remain findable, accessible, interoperable and reusable (Wilkinson et al., 2016). Students also discuss obligations to local communities, landowners and heritage authorities. A practical task is to prepare a mock permit application for a defined area, including a budget, a timetable and a risk register.

3.3. Stage 3: Multidisciplinary Investigation

Stage 3 is the analytical core. Several disciplines examine the same problem using their own methods while remaining independent of one another. The evidence is organized into four streams. In teaching, each stream becomes a training module. Every module ends with the same task: compare the result with an explicit benchmark and state what the comparison does and does not show. Table 2 lists the modules.

Table 2. The four evidence streams of Stage 3 as training modules.

Note: The values in the last column are results from the research program and are used as teaching material. They are open to critique in class and are not presented here as settled findings.

3.3.1. Stream 3A: Archaeology, Laboratory Analyses and Chronology

Excavation establishes context: stratigraphy, features, artifacts and deposits (Harris, 1989). Laboratory work then characterizes materials and dates. Chronology should rest on more than one dating method where possible (Aitken, 1990). Samples should follow an explicit sampling design, so that their representativeness can be judged (Banning, 2002; Orton, 2000). Geoarchaeological analysis of sediments links the two (Goldberg & Macphail, 2006).

The compressive-strength results from Visoko provide a useful exercise in benchmark choice. The reported values were compared with modern concrete. Students can ask whether that comparison answers the research question. If the question is whether the material is natural or artificial, then cemented conglomerates and sandstones sampled from local natural outcrops may be a more informative control than modern concrete. The exercise teaches that choosing a benchmark is itself a methodological decision. It must be justified, not assumed.

The same point applies to laboratory testing in general. The program reports that samples from the site have been tested at construction and materials institutes in Bosnia and Herzegovina, the Czech Republic, Slovakia, Italy and France (Osmanagich, 2025a). Such tests are not limited to measurement quality. Mineralogy, petrography, binder chemistry and microstructure can carry information on whether a material was manufactured. Their power to discriminate, however, depends on the comparison. The same tests must be applied, under the same protocol, to natural rock of the same lithology from the surrounding area. Accredited laboratory procedures make a value reliable; the geological comparator is what turns a reliable value into a test of origin.

3.3.2. Stream 3B: Geodesy, GIS and Remote Sensing

Precise spatial documentation is the basis for any landscape analysis (Conolly & Lake, 2006; Wheatley & Gillings, 2002; Kvamme, 1999). Airborne laser scanning can reveal features even under forest cover (Doneus et al., 2008). Within the framework, geospatial results are first treated as descriptions, not as inferences. A regular shape or a precise orientation is a measured variable. It is not yet evidence of design.

The reported orientation of a hill face close to true north is a good classroom case. Students can be asked how to test such a value. One approach is to measure the orientation of many natural ridges and faces in the same region, using the same DEM and the same definition of a “face”. The observed value is then placed within that distribution. Students learn that a precise measurement is not the same as an improbable one.

3.3.3. Stream 3C: Geophysics and Environmental Measurement

Non-invasive methods such as GPR, resistivity and magnetometry examine the subsurface without disturbing it (Clark, 1996; Gaffney & Gater, 2003). Their results are anomalies: departures from an expected background. An anomaly is an observation, not an interpretation. Its meaning has to be established by excavation, coring or comparison with known natural structures.

Students in this module design a control. If measurements are taken on a hill of interest, the same instruments, settings and protocols must be applied to natural hills of similar geology and size. Students also learn to report the number of anomalies relative to the area surveyed and to the background rate on control sites. Without a control, a count of anomalies has no meaning.

3.3.4. Stream 3D: Spatial Statistics and Archaeoastronomy

Apparent alignments and geometric patterns are common in any landscape with many features. They must be tested against explicit null models (Baxter, 2003; Bevan & Lake, 2013; Bivand et al., 2013). Archaeoastronomy faces the same problem. With many candidate features and many possible astronomical targets, some alignments will occur by chance (Ruggles, 2015).

This module teaches simulation and the problem of multiple testing. Students build a Monte Carlo test for a simple alignment hypothesis. They then repeat it after examining many candidate patterns and see how the false-positive rate changes. The lesson connects to current debates on reproducibility and preregistration (Munafò et al., 2017; Nosek et al., 2018). One Visoko study addresses this problem directly: hypotheses formed after inspecting the data (Osmanagich, 2026b). Students can read it critically and ask whether its constraints are sufficient. A further principle is stated explicitly: statistical significance shows that a pattern differs from a null expectation. It does not establish cultural meaning.

3.3.5. Integration across Streams

The final task of Stage 3 is integration. Students build an evidence table. Each row is one finding; the columns record the stream, the benchmark, the result of the comparison, the alternative explanations still open and the degree of independence from the other findings. Two results that depend on the same measurement do not count as two lines of evidence. The table makes this visible. It also shows where streams converge, where they conflict and where evidence is simply missing.

Table 3 gives a worked example. It uses one reported finding from each stream of the Visoko program and records it in the format students are asked to use. The entries show how the framework treats a result before any interpretation is attached to it.

Table 3. Worked example of a completed evidence table, using reported findings from the Visoko program.

Note: The entries illustrate the format and the standard the framework applies. The last column is a provisional judgment within the framework, not a final verdict on the site. In class, students complete a parallel table for the natural hypothesis.

Two lessons follow from the example. A strong-looking number may still lack the control that would make it discriminating. And four findings from four streams may amount to fewer than four independent lines of evidence.

3.4. Stage 4: Scientific Evaluation

Research is not complete when the data are analyzed. Results must be exposed to colleagues who did not produce them. Conferences, workshops and invited lectures provide that exposure. The aim is not agreement but scrutiny. Criticism identifies weak methods, missing controls and alternative explanations. Over two decades, results from the Bosnian program have been presented at international meetings and have drawn both support and strong criticism. The program has also organized nine International Conferences on Bosnian Pyramids (ICBP) between 2008 and 2026 (Section 7).

The educational function of Stage 4 is to present, debate scientifically, and provide peer feedback. Students present their evidence tables to a panel and to classmates. One group is assigned a structured role: to argue the strongest case against the presenters’ interpretation. Collaborative critical discourse of this kind improves scientific reasoning (Osborne, 2010). Students also give written peer feedback using a shared rubric. Peer assessment has been shown to support learning when criteria are clear and feedback is specific (Topping, 1998; Nicol & Macfarlane-Dick, 2006).

3.5. Stage 5: Peer-Reviewed Publication

Peer review does not guarantee that a conclusion is correct. It provides an independent check by specialists on whether methods, analyses, and interpretations meet accepted standards (Day & Gastel, 2012). Publication also makes methods and data permanent and open to replication. The author’s online bibliography lists 59 scientific articles from the program as of October 2026, with the title, journal and a link for each (Osmanagich, 2026c). They cover excavation, chronology, geodesy, geophysics, laboratory analyses, environmental measurement and spatial statistics. Readers can identify each work and its venue there.

The educational function of Stage 5 is academic writing, transparency, and communication. Students write a short article in journal format. They also act as reviewers for one another, using a standard review form. Marwick (2017) has shown how reproducible workflows, with shared code and data, can be built into archaeological research. Students are asked to deposit their data and scripts with their manuscript. The final question of Stage 5 is the practical test of transparency: could another team repeat the work from what has been written and shared?

4. Educational Applications of the Five-Stage Framework in Higher Education

The framework can be used at several levels of training. The structure stays the same. The depth, autonomy, and size of the tasks change.

4.1. University Courses

At the undergraduate level, the framework can organize a single methods course in archaeology, physical geography, or environmental science. Each stage becomes a block of teaching. Students work in small interdisciplinary teams, ideally mixing students of archaeology, geodesy, geology and statistics. Each team takes one landscape question from hypothesis to a short paper. Published datasets, open LiDAR data and open satellite imagery allow much of the work to be done without fieldwork.

This design is close to course-based undergraduate research experiences (CUREs). In a CURE, an entire class participates in research with a real, unknown outcome (Auchincloss et al., 2014). The five-stage framework gives such a course a clear sequence and a shared standard of evidence.

At master’s level, the framework can serve as a capstone. Students design their own project, carry out at least two of the four streams with real data, and present the results at a student conference organized as Stage 4.

4.2. Field Schools

Field schools are the setting closest to real research. They often concentrate on excavation technique. The framework widens their scope. Before fieldwork, students write hypotheses and a research design. During fieldwork, they rotate through the four streams: excavation and sampling, survey and photogrammetry, geophysics, and spatial analysis. Each rotation ends with a benchmark exercise. After fieldwork, students present results in a closing seminar and prepare a short report.

The Visoko program has run field and volunteer seasons over many years, in which participants work on excavation, documentation, and tunnel clearing. These seasons show that the logistics of a multidisciplinary field school are workable at such a site. They do not show that the five-stage model improves learning. That remains to be tested (Section 6).

4.3. Training Early-Career Researchers

For doctoral students and young researchers in research centres, the framework works as a checklist for project design and as a structure for supervision. Supervisors can review a project stage by stage. Has the hypothesis been stated with its competing alternatives? Is the program authorized and documented? Does each finding have a benchmark? Has the work been presented to critics before submission? Is the manuscript reproducible? A short written review at the end of each stage gives the candidate a record of progress and gives the supervisor an early warning of weak points.

4.4. Using a Contested Site as a Teaching Case

Contested sites are valuable in teaching because they make the rules of evidence visible. When a claim is uncontroversial, students rarely ask how it was established. When a claim is disputed, they must examine the evidence, the controls, and the reasoning on both sides. Feder (2025) uses disputed claims this way in a widely used textbook on the scientific method in archaeology.

The Visoko case can be taught as a structured debate. Students receive primary publications from the research program, published critiques and an independent geological reading of the site. The geological reading comes from a team of the Faculty of Mining, Geology and Civil Engineering at the University of Tuzla. In 2006 the team drilled six boreholes, 3 - 17 m deep, on Visočica. It described the hill as a natural formation of layered clastic sediments (conglomerate, sandstone and clay) of Miocene age, shaped by tectonic and erosional processes, and noted similar landforms elsewhere in the Sarajevo-Zenica basin (Vrabac et al., 2006). This report gives students the natural-landform explanation in the words of geologists, not only as an alternative named by the program. They build evidence tables for both the anthropogenic and the natural hypotheses. They identify which observations each side has made, which controls are missing, and what new observation would most reduce the uncertainty. The goal is not to reach a verdict in class. The goal is to practice judging evidence fairly, including evidence for a view the student does not share.

5. Learning Outcomes and Competencies

Table 4 links each stage to intended learning outcomes, teaching activities and assessment methods. The outcomes are written at the level expected at the end of a master’s program. For undergraduate courses, the same outcomes can be set at a lower level of independence.

Table 4. Competency matrix: learning outcomes, teaching activities and assessment for each stage.

Note: Outcomes progress from the “analyze” and “evaluate” levels at Stages 1 - 3 to the “create” level at Stage 5 of the revised Bloom taxonomy (Anderson & Krathwohl, 2001).

Three competencies cut across all five stages. The first is benchmark reasoning: the habit of asking “compared with what?” before interpreting any result. The second is interdisciplinary integration: the ability to combine results from different fields while tracking their independence and their limits. The third is intellectual honesty: reporting null results, stating limitations, and representing opposing views fairly. These three are the main targets of the model. They are also the hardest to assess, which is why each appears in more than one assessment task.

6. Implementation and Evaluation

6.1. Proposed Curriculum

Table 5 outlines a one-semester course of fifteen weeks, worth 6 ECTS credits, followed by an optional two-week field module worth a separate 3 ECTS credits. In the European Credit Transfer and Accumulation System, one credit corresponds to 25 - 30 hours of total student work (European Commission, 2015). The 6-credit course therefore requires 150 - 180 hours: 45 contact hours (three hours per week for fifteen weeks) and 105 - 135 hours of independent and team work. The field module comprises ten field days of about 7 hours each, plus preparation and a closing report, for a total of 75 - 90 hours, or 3 credits. Students who take both components earn 9 credits. The course is designed for advanced undergraduate or first-year master’s students in archaeology, geodesy, geology, geography or environmental science. Teams of four to five students should include at least two disciplinary backgrounds.

Table 5. Proposed curriculum: one-semester course (15 weeks, 6 ECTS) with optional field module (2 weeks, 3 ECTS).

Note: 1 ECTS credit = 25 - 30 hours of student work (European Commission, 2015). Weeks can be shortened or merged for undergraduate courses. Where fieldwork is not possible, open datasets can replace the field module; the course then carries 6 ECTS only.

The course can be taught by a team of instructors from different disciplines. Where that is not possible, guest sessions from specialists can cover individual streams. The structure of the framework keeps the course coherent even when teaching is shared.

6.2. Assessment

Assessment follows the principle of constructive alignment (Biggs, 1996; Biggs & Tang, 2011). Each major task is scored with an analytic rubric shared with students in advance. Rubrics improve scoring reliability and help students understand expectations (Jonsson & Svingby, 2007). A suggested weighting is: hypothesis and research design 15%; stream practicals and benchmark reports 25%; evidence table 20%; conference presentation and peer feedback 15%; final manuscript and reproducibility check 25%.

Formative feedback is given at the end of every stage, before the summative tasks. Students revise their work in response. This cycle mirrors the research process, in which results return to earlier stages for revision. It also supports self-regulated learning (Nicol & Macfarlane-Dick, 2006; Black & Wiliam, 1998).

6.3. Proposed Design for Empirical Evaluation

The claim that the framework improves learning has not been tested. It should be tested by researchers who are independent of the Visoko program. Table 6 outlines a possible design.

Table 6. Proposed design for future empirical evaluation of the educational model.

Note: The design should be preregistered before data collection. Ethical approval and informed consent are required.

A randomized design is rarely possible in university teaching. A quasi-experimental design with a comparison group and pre-tests is a practical alternative, provided its limits are stated (Shadish et al., 2002). Raters should not know which group a student belongs to. The study plan, hypotheses and analyses should be preregistered (Nosek et al., 2018). Evaluation can also follow Kirkpatrick’s levels, from student reactions to learning, behavior, and results (Kirkpatrick & Kirkpatrick, 2006). The behavior level, whether graduates use benchmark reasoning in their later work, is the most important and the hardest to measure.

Courses are rarely comparable in every respect, and three differences matter most when results are interpreted. The first is students’ disciplinary preparation. A class with many geodesy or statistics students may score higher on integration tasks for reasons unrelated to the framework. The evaluation should record each student’s degree program and prior methods courses and use them, with pre-test scores, as covariates. The second is instructor expertise. If the framework course is taught by more experienced instructors, a gain may reflect the teachers rather than the model. Where possible, the same instructors should teach both versions, in parallel sections or alternating years; otherwise their experience should be recorded and reported. The third is teaching time. The comparison course should carry the same credits and contact hours, and both courses should log the time spent on each topic. When several courses and instructors are involved, students should be analyzed as nested within course sections, using multilevel models (Raudenbush & Bryk, 2002), so that a difference between sections is not mistaken for an effect of the model. Differences that remain after these adjustments should still be attributed to the model with caution.

Replication at several institutions would be the strongest test. If the model works only at the site where it was developed, or only with its authors as teachers, that would be an important result in itself.

7. Discussion

The framework does not introduce new methods of research or teaching. Its contribution is organizational. It places existing methods in a sequence, adds one rule that applies at every step and pairs each step with explicit learning outcomes. The benchmark rule is the core of the model. It is simple enough to teach in the first week and demanding enough to challenge experienced researchers.

The framework’s distinguishing features are set out in Section 2.5. It extends existing inquiry and research-based models in three ways: 1) it carries communication and reflection beyond the classroom, into legal organization, external evaluation and publication; 2) it requires a justified benchmark for every result, and separates calibration against reference standards from discrimination against natural comparators; and 3) it trains students to separate independent lines of evidence from several analyses of the same data. Existing models address parts of the first point. They rarely make the second and third explicit.

The framework also makes visible parts of research that students often overlook. Permits, archives, data plans, conference criticism, and peer review rarely appear in methods courses. Yet they decide whether research can be trusted. Teaching them as stages of the scientific process, not as administration, gives students a more realistic picture of science.

Using the Bosnian Valley of the Pyramids as the case study requires a clear statement, with dates. In 2006, after a field visit, Anthony Harding, then president of the European Association of Archaeologists, concluded that the formations he examined were natural (Harding, 2007). In the same year, a geological team from the University of Tuzla reached the same conclusion for Visočica from borehole data (Vrabac et al., 2006). These assessments came at the very start of the program, before most of the later work. Nearly two decades later, the 11th edition of a widely used methods textbook still discusses the site as a case of disputed archaeological claims (Feder, 2025). The interpretation of the Visoko hills as anthropogenic constructions is therefore not accepted, as of 2025, in the published positions of mainstream archaeology and geology cited here.

The research program has continued over the same period. It has organized nine ICBP conferences (2008, 2011, 2012, 2013, 2014, 2018, 2024, 2025 and 2026). According to the organizers, more than 100 scientists from a range of disciplines and countries have taken part (ICBP, 2026). The program has also published laboratory, geodetic, geophysical and statistical results, some of them written as direct responses to criticism; one recent study, for example, deals specifically with the risk of hypotheses formed after inspecting the data (Osmanagich, 2026b). Table 7 lists selected milestones, with their limitations and the critiques that apply to them.

Table 7. Selected chronology of the Visoko research program, 2005 - 2026, with limitations and critiques.

Note: The table lists results reported by the program and the critiques cited in this paper. It is not a complete bibliography of the program or of its critics. A dash means that no published independent assessment specific to that item is cited here.

Two points follow. First, critiques and results need dates. An early critique does not settle questions raised by later measurements, and later publications do not cancel an early critique unless they answer it directly. Second, publication of results, including in peer-reviewed journals, does not by itself confirm an anthropogenic origin. Conferences organized by the program do not replace evaluation by independent bodies. This paper does not try to settle the dispute. The framework was derived from the program’s experience, but its value does not depend on the program’s conclusions. Its rules would apply equally to a team testing the natural hypothesis. For teaching, the dispute is an asset. It gives students a real case, with a documented history, in which the quality of controls, benchmarks and statistical tests matters, and in which they must weigh arguments from both sides.

Several limitations should be stated. First, the educational model has not been evaluated. All statements about its benefits are expectations based on the wider literature on research-based and interdisciplinary learning. Second, the model places heavy demands on instructors. Teaching four evidence streams well requires expertise or collaboration across departments. Third, real fieldwork is expensive. Open datasets reduce this cost but cannot fully replace the experience of a field season. Fourth, the framework was developed at a single site by a single research program. Its usefulness elsewhere, in research and in teaching, remains to be shown. Finally, convergence among methods raises confidence but does not remove the need for continued testing as new evidence appears.

8. Conclusion

This paper has proposed a five-stage framework for teaching interdisciplinary scientific research, built on archaeology and geospatial technologies. The stages are hypothesis formulation, research organization, multidisciplinary investigation, scientific evaluation, and peer-reviewed publication. Each stage has a scientific function and an educational function. A single rule connects them: every finding must be compared with an explicit benchmark before it is interpreted.

The paper describes how the framework can be applied in university courses, field schools, and research training. It sets out learning outcomes, teaching activities, and assessment methods for each stage, a one-semester curriculum, and an empirical evaluation design.

The framework draws on two decades of multidisciplinary research in the Bosnian Valley of the Pyramids. That program is the source of the model and of much of its teaching material. It is not evidence that the model works in education. Independent teams must now test the framework’s educational value at different institutions and sites. The model is offered as a starting point for that work.

Conflicts of Interest

The author founded and led the research program used as the case study in this paper. The educational model has not been evaluated by the author or by others, and no student data are reported.

References