Convergence Methods

Core Concepts

Triangulation as an Epistemic Standard

No single strand of unwritten evidence can do much on its own. An oral tradition without corroborating physical evidence is vulnerable to the charge of myth-making. A genetic cluster without an archaeological counterpart leaves population movements underdetermined. A landscape reading without independent textual or oral support risks projecting the observer's categories onto the past.

Triangulation addresses this by requiring that a finding be affirmed through multiple independent strands of information. The concept is explicit in the oral history and archaeology literature: triangulation increases credibility and validity precisely because independent convergence is difficult to produce by chance or bias alone.

Independence is the operative word. Evidence only triangulates if the methods generating it carry distinct error profiles — otherwise, convergence merely reveals a shared assumption.

The three key properties for genuine triangulation are:

1. Methodological independence — the methods must not share data sources, instruments, or interpretive frameworks.
2. Epistemic complementarity — each stream should cover different aspects of the claim, filling in what the others cannot reach.
3. Temporal or procedural separation — the findings should have been developed without knowledge of one another where possible, to rule out circular confirmation.

What Counts as Evidence?

A productive shift in recent historiography has been the expansion of what counts as a legitimate "archive." Multispecies historiography argues for treating natural archives — ice cores, lake sediments, pollen cores, tree rings, isotope records — as equivalent to textual sources, woven into "geospatial-narrative archives" alongside oral testimony, archaeological evidence, and written documents.

Paleoenvironmental proxies such as dendrochronology (tree-ring dating) and oxygen isotope stratigraphy provide quantifiable records of past climatic and ecological conditions. These proxies reconstruct vegetation shifts, drought cycles, and human environmental impact — often the only evidence available for periods and places without written records.

The implication is methodological: a historian working with unwritten sources now has access to a toolkit far wider than oral testimony and artifacts alone. This breadth, however, also multiplies the points at which assumptions can slip in unnoticed.

Negative Evidence and Survivorship Bias

One of the subtler traps in multi-method work is misreading the absence of evidence as evidence of absence. This conflation is a specific form of survivorship bias: the tendency to concentrate analysis on what survived a selective filter while systematically ignoring what did not.

Survivorship bias is formally defined as a selection bias that results from focusing on entities that passed a filter — and it is self-concealing, because what failed to survive disappears from the dataset by definition. In historical research, this means:

• The records that survive are not a random sample of the past; they are a sample of what was preserved, which is shaped by power, accident, material durability, and institutional interest.
• When a method finds nothing, that null result may reflect the limits of the method, the conditions of preservation, or access constraints — not the actual absence of the event or practice being sought.

Publication bias compounds the problem at the level of scholarship: studies with positive or significant findings are more likely to be published, leaving null results in the "file drawer." This means the available literature skews toward successful convergences, understating how often multi-method approaches simply find nothing.

Negative evidence only counts when you can specify what would have been preserved had the event occurred — and demonstrate that preservation conditions were adequate. Without that, absence is not data; it is silence.

Microhistory as a Bridge to Macrohistory

A recurring problem in multi-method work is scale: individual cases and population-level patterns use different kinds of evidence and support different kinds of inference. Osteobiography as microhistory offers a useful model for navigating this.

Osteobiographical analysis combines skeletal morphology, isotope studies, ancient DNA, archaeological context, and archival sources to reconstruct individual life histories. But recent scholarship insists this microscale work must be in constant dialogue with population patterns — moving between the individual case and macroscale historical themes. Findings about a single individual only gain historical significance when situated against a population baseline; conversely, population-level claims require validation from individual cases that exemplify or deviate from them.

This bidirectional movement — from micro to macro and back — is a general principle applicable whenever you are integrating different evidential scales.

---

Step-by-Step Procedure

A Framework for Evaluating Multi-Method Historical Claims

This procedure applies when you encounter a claim supported by more than one type of evidence and need to assess whether the convergence is meaningful.

Step 1 — Inventory the evidence streams. List every type of evidence cited. Classify each by its method: oral tradition, genetic analysis, archaeological stratigraphy, paleoenvironmental proxy, material culture, linguistic phylogenetics, written record. Note the source for each.

Step 2 — Test for genuine independence. For each pair of evidence streams, ask: do they share data sources, interpretive frameworks, or researchers? If two streams derive from the same fieldwork, use the same assumptions about cultural transmission, or were developed by the same team aware of each other's results, they do not fully triangulate. As established in the genetics-archaeology literature, genetic, archaeological, and linguistic evidence each carry independent sources of error — genuine convergence requires that independence to be preserved.

Step 3 — Map the error profiles. What can each method get wrong? Oral traditions face reliability challenges at different levels depending on whether you are dealing with individual memory (reconstruction effects, recall bias) or transmission chain dynamics (attrition, deliberate reshaping). Genetic analysis can mistake sampling gaps for population boundaries. Landscape phenomenology faces the verification irreproducibility problem: the archaeologist's embodied reading of a site cannot be reproduced or falsified by another reader. Archaeoastronomy must contend with false positives — alignments that arise by chance, detectable only through Monte Carlo simulation or independent corroboration.

Step 4 — Check for shared biases. Even independent methods can share systematic biases. Survivorship bias in the archive means every method is more likely to recover evidence about powerful, stable, or institutionally backed groups. Ask: does each stream sample from the same selective filter? If so, their convergence may reflect a shared gap rather than a shared truth.

Step 5 — Evaluate the null result explicitly. If one or more methods found nothing, assess whether that null result is informative. What would you expect to find, given adequate preservation conditions and method sensitivity? If the answer is "something," then absence counts against the claim. If the answer is "we don't know," the null result is uninformative and should not be treated as confirming or disconfirming evidence.

Step 6 — Assess scale coherence. Do the evidence streams operate at compatible scales? A claim about a population migration should not rest primarily on a single individual's isotope data; a claim about a named historical figure should not be resolved by genetic averages. Use the microhistory-to-macrohistory principle: individual cases and population patterns must validate each other bidirectionally, not simply be stacked together.

Step 7 — State your confidence level and its basis. Distinguish between: (a) convergence that is strong because streams are genuinely independent and have different error profiles; (b) convergence that is provisional because some independence conditions are uncertain; (c) apparent convergence that may reflect shared assumptions. Be explicit about which type you have.

---

Worked Example

The 1700 Cascadia Earthquake

The 1700 Cascadia earthquake is the benchmark case for convergent evidence across independent epistemological domains. It is worth walking through in detail because it demonstrates what genuine triangulation looks like — and what made confidence so high.

The three streams:

1. Oral traditions preserved by First Nations communities along the Pacific Northwest coast described a nighttime catastrophic earthquake and tsunami. These accounts were passed down across generations and collected by ethnographers.

2. Paleoseismic evidence documented ghost forests, coastal subsidence, and buried soil layers — physical signatures of sudden coastal drop consistent with a megathrust rupture. Dendrochronological death dates from the ghost trees pointed to a range of 1699–1700.

3. Japanese written records documented an orphan tsunami on January 27–28, 1700 — a tsunami with no local earthquake to explain it, arriving from across the Pacific.

Why this constitutes genuine convergence:

Each stream was developed independently using distinct methods. The oral accounts use ethnographic and linguistic analysis. The paleoseismic record uses sediment stratigraphy, geomorphology, and dendrochronology. The Japanese tsunami records use documentary history and tidal gauge interpretation. None of these share data sources, instruments, or interpretive frameworks.

Each stream also carries distinct error types. Oral accounts may compress, mythologize, or conflate events across generations. Paleoseismic dating carries uncertainty ranges and depends on the integrity of the sediment column. Japanese documentary records are subject to scribal error and institutional selection. Because the errors are different, the chance of all three streams producing a false positive for the same date is extremely low.

The result is remarkable: all three lines converge on January 26, 1700 — the date of the megathrust rupture. The oral traditions' description of a nighttime event aligns with the inferred rupture time on the Pacific Northwest coast, and the travel time of the resulting tsunami matches the arrival times in Japanese records.

The lesson:

The Cascadia case demonstrates the full value of triangulation: when independent streams not only agree on an event but also agree on details (time of day, tsunami reach) that neither stream alone could determine, confidence rises substantially. It also shows what oral tradition can contribute when treated as evidence: the First Nations accounts were not simply "confirmed" by science — they helped calibrate the paleoseismic dating and informed the search for the tsunami source.

---

Boundary Conditions

When Convergence Breaks Down

Shared-assumption convergence. When multiple methods draw on the same underlying assumptions, their agreement tells you the assumption is consistent — not that it is correct. In archaeoastronomy, both statistical and corroborating-evidence schools emerged precisely because raw alignment counts can produce false positives that look like convergence. If you run the same theoretical model through three different datasets, you are not triangulating the model — you are replicating it.

Irreproducibility barriers. Some methods generate results that cannot, in principle, be independently verified. Phenomenological landscape archaeology exemplifies this: the archaeologist's embodied experience of a site is the primary data, but as Brück (2005) argues, no reader can determine whether the relationships identified in the present were considered significant in the past. Two trained researchers can produce mutually incompatible readings of the same landscape, both defensible within the framework. Methods with this property cannot serve as independent triangulation legs; they can only be one interpretive layer among others with external verification hooks.

Scale mismatch. Evidence from different scales does not automatically integrate. A genetic signal averaged over a population tells you something different from an isotope record from a single skeleton. Osteobiography specifically addresses this by requiring constant movement between individual variation and population patterns. Claims that simply aggregate evidence across scales without explaining how individual cases relate to population averages — or vice versa — are not genuinely multi-method; they are multi-scale without integration.

Differential preservation masquerading as pattern. Survivorship bias can produce apparent convergence by ensuring that all surviving evidence comes from the same kinds of contexts: sites that were inhabited continuously, groups that had political power, practices that left durable physical traces. If all your evidence streams preferentially recover the same subset of the past, their convergence shows you what is preserved — not what happened. Dress history illustrates this: object-based methodology combining surviving garments, inventories, sumptuary laws, and paintings inevitably skews toward elite dress, because elite dress survived. Agreement across those streams does not validate a claim about dress in general.

Transmission chain distortion vs. individual memory error. When using oral traditions alongside oral history, the reliability challenges operate at different levels and in different ways: oral traditions face attrition, deliberate reshaping, and censorship across transmission chains, while oral history faces reconstruction effects at the individual cognitive level. Treating both as "oral evidence" and assuming they carry the same error profile will produce a misleading picture of how robust the oral component of your triangulation actually is.