Inverted Dates at the Rockshelter

Anomalies to explain: (a) Layer 2 older than Layer 3 beneath it; (b) two charcoals in one layer differing by ~900 years; (c) TL at 1,800 BP versus a 2,150 BP charcoal in the same layer.

Ranked explanations:

1. Old-wood / inbuilt age - best explains (b). Charcoal from a long-lived hardwood gives the age of the inner rings, not of the burning event. A 900-year spread within one layer is textbook. It also weakens (a), since the Layer 2 charcoal may simply be dead wood burned late.
2. Intrusive charcoal via the rodent burrow - best explains (a). The burrow is drawn cutting from Layer 2 into Layer 4; charcoal and sediment can migrate downward, making an upper layer appear older or a lower layer appear younger. The root channel on the east face does the same vertically.
3. Redeposition / inverted stratigraphy - competing explanation for (a). Slope wash in a rockshelter can rework deposits; the section drawing must be checked for a relict cut-and-fill rather than assumed to be horizontal.
4. Comparing unlike events - best explains (c). TL dates the last firing of the ceramic, i.e. manufacture; a curated or heirloom pot can be deposited centuries later. Conversely, the charcoals may date wood older than the occupation. Two ages for one layer are not automatically contradictory; they date different events in the object's life.
5. Laboratory or calibration error - least likely as a first resort. Both samples carry stated errors of 60–90 years; the discrepancy is roughly ten sigma, so analytical error is insufficient unless sample pretreatment failed (e.g. carbonates, humic contamination), which is worth checking.
6. Disturbance by later pit-digging, ash dumping or bioturbation by termites - plausible in tropical shelters and consistent with the burrow evidence.

Point to stress: (a) and (b) may share a single cause (old wood plus intrusion), so the anomalies should not be counted as independent evidence of a 'complex sequence'.

Radiocarbon on charcoal. Dates the death of the wood, minus any inbuilt age. It is expressed in uncalibrated ¹⁴C years 'BP' (before AD 1950) and must be calibrated (IntCal curve, Bayesian software such as OxCal) to give calendar years. Furthermore, a conventional bulk charcoal date averages all fragments in the sample; AMS on a single identified short-lived item (a charred millet or rice grain, a twig, resin) is required to date a burning event rather than a forest.

Thermoluminescence on pottery. Dates the last time the ceramic minerals were heated - firing in a kiln or hearth. Deposition may follow much later, and the sherd may be residual or curated. TL also requires dosimetry at the find spot, which is why a sherd analysed out of its sediment context has limited value.

OSL on sediment. Dates the last exposure of quartz or feldspar grains to light, thereby dating deposition of the enclosing sediment. It is the direct check on whether Layer 3 is genuinely in place - not yet used here and arguably the most valuable missing measurement.

Why comparison fails as quoted.

• Uncalibrated ¹⁴C cannot be read as calendar years, and the calibration curve is not a constant offset.
• Two sigma ranges overlap for some pairs but not others; a bare ± figure with no laboratory code, pretreatment and δ¹³C is not a date.
• TL and ¹⁴C date different events, so disagreement is not necessarily error.
• '2900 BP' cannot be translated into 'iron-using communities' without a dated association between slag and that charcoal in a sealed context.

Rule to state in the report: no chronological claim at this site may be made without a calibrated range, a laboratory code, and an explicit statement of which event is dated.

Permitted wording (final abstract). 'The rockshelter contains microlithic occupation and later ceramic-using and iron-working activity. Charcoal from the upper levels indicates activity in the later second to early first millennium BC in calibrated terms, but the sequence is disturbed by a rodent burrow and a root channel, and two assays from one layer differ by c. 900 radiocarbon years, indicating old-wood or intrusive material. No secure date for the first iron use at the site is yet available.'

Why this wording. It reports what the samples can bear: presence, a broad calibrated horizon, and the existence of a taphonomic problem. It deletes 'unbroken continuity', which a burrow alone contradicts, and deletes 'iron-using communities from 2900 BP', which requires a tight association between slag and a dated short-lived sample.

Resampling and analysis plan.

1. Re-excavate the north face to record the burrow fill separately; sample the burrow fill as its own context.
2. Collect 8–10 single-entity AMS samples (identified short-lived taxa) from Layers 2 and 3, avoiding root channels.
3. Submit at least two paired charcoal samples with ABA/humic-acid pretreatment to test contamination.
4. Take OSL tubes from Layers 2 and 3 to test whether the sediment is in place.
5. Collect TL samples in situ with dosimetry for any further ceramic dating.
6. Run a Bayesian stratigraphic model with the burrow as an intrusion, and publish the model, not a single 'best' date.

Publication split. The lithic and ceramic description and the taphonomic note can go now; the chronology and the iron-first-use claim wait for the new series.