The DOI matches its Crossref record; 49 of the 52 references it cites were found in a registry or an index; 5 of the 27 sources its claims cite were checked for support.
Explaining Known Past Routes, Underdetermination, and the Use of Multiple Cost Functions
Open accessAbstract
Explaining material traces of movement as proxies for past movement is fundamental for understanding the processes behind why people in the past traversed the landscape in the way that they did. For this, least-cost path analysis and the use of slope-based cost functions for estimating the cost of movement when walking have become commonplace. Despite their prevalence, current approaches misrepresent what these cost functions are, their relationship to the hypotheses that they aim to represent, and their role in explanation. As a result, least-cost paths calculated using single cost functions are liable to spurious results with limited power for explaining known past routes, and by extension the decision-making processes of past people. Using the ideas of multiple model idealisation and robustness analysis, and applied via a tactical simulation, this study demonstrates that similar least-cost paths can be produced from slope-based cost functions representing both the same hypothesis and different hypotheses, suggesting that least-cost path results are robust but underdetermined under the tested environmental settings. The results from this tactical simulation are applied for the explanation of a Roman road in Sardinia. Using probabilistic least-cost paths as an approach for incorporating multiple cost functions representing the same hypothesis and error in the digital elevation model, it is shown that both model outcomes representing the minimisation of time and energy are unable to explain the placement of the Roman road. Rather, it is suggested that the Roman road was influenced by pre-existing routes and settlements.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
Using multiple model idealisation and robustness analysis, this paper argues that least-cost path results from slope-based cost functions are often robust but underdetermined: different hypotheses can produce similar paths. A tactical simulation and a Roman road case study in Sardinia show that time- and energy-minimisation hypotheses cannot reliably explain known past routes, with pre-existing routes and settlements suggested instead.
Key findings
- In a tactical simulation using five DEMs (1 km × 1 km, 1 m resolution; fractal dimensions 2.20–2.60) and two elevation scenarios (0–5 m and 0–10 m), deviation within and across time- and energy-based cost functions was small. The hypotheses were therefore robust but underdetermined across all landscape complexities and scenarios tested.
- The simulation used 14 cost functions and calculated 91 Euclidean distance values per run; repeated 1000 times, this produced two sets of 455,000 least-cost paths (as reported: 91 LCPs × 5 landscapes × 1000 simulations per scenario). Increasing Euclidean distance between origin and destination increased deviation within and across hypotheses, but the two hypotheses remained underdetermined.
- For the ‘a Karalibus Sulcos’ Roman road in south-west Sardinia, probabilistic least-cost paths using multiple time- and energy-based cost functions showed that neither minimising time nor minimising energy could explain the road’s placement. Some energy-based paths followed the road, but only the Herzog (2014c) cost function produced them, suggesting model idealisation rather than the shared hypothesis.
- The paper suggests the Roman road was influenced by pre-existing routes and settlements, including possible Nuragic transhumance routes and Punic-era roads, rather than by minimising time or energy on foot.
Methods
- Case study 1 was a tactical simulation using five synthetic DEMs generated with the spectral synthesis method, scaled to 0–5 m and 0–10 m, with fractal dimensions of 2.20, 2.30, 2.40, 2.50, and 2.60. Fourteen time- and energy-based cost functions were used to calculate least-cost paths between random origin and destination points, with pairwise maximum Euclidean distances used to assess robustness and underdetermination.
- Case study 2 applied probabilistic least-cost paths to the ‘a Karalibus Sulcos’ Roman road in Sardinia, using the TINITALY 10 m resolution DEM (RMSE 4.3 m). Fourteen time- and energy-based cost functions were used concurrently, with each LCP calculated 50 times under different realisations of random DEM error.
- Least-cost paths were calculated using the Dijkstra algorithm and the leastcostpath R package; the tactical simulation used a 4-adjacency neighbourhood, while the Sardinian case study used a 16-adjacency neighbourhood.
Limitations
- The paper states that the tactical simulation is topography-only and does not include rivers or different land types; it also excludes the Pandolf et al. (1977) energy cost function because it requires body mass, load mass, and terrain quality often unavailable in archaeological contexts.
- The paper states that rivers were not incorporated into the Sardinian case study because it is difficult to know which were passable or required fords or bridges; the Roman road was digitised by the author because a digitised version was not available; and dating settlements along the road is difficult.
- The paper states that autocorrelation in random DEM error was not accounted for, using unfiltered error fields as a worst-case scenario; it also notes that DEMs may contain modern-day roads, and that similarity between paths is subjectively defined by the analyst. The paper’s own conclusion notes that underdetermination limits the explanatory power of the time- and energy-minimisation hypotheses.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
5 of 27 cited sources checked · 2 supported · 0 flagged · accuracy not yet measured
- Premisecitation
The 'best' cost function from Güimil-Fariña and Parcero-Oubiña (2015) was subsequently used in Fonte et al. (2017).
“The 'best' cost function was subsequently used in Fonte et al. (2017)”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites Fonte et al. (2017) source
Section 3.1 says Fonte et al. followed Güimil-Fariña and Parcero-Oubiña’s methodology and used the Llobera–Sluckin cost function in the analyses they present. It also says Fonte et al. tested the functions for their own study area; our copy does not identify a single function as the earlier study’s ‘best’ one. (LLM, conf=0.82)
Experimental — accuracy not yet measured
- Study design and conductcitation
Autocorrelation in random error was not accounted for, with completely random unfiltered error fields used instead.
“In contrast to Lewis (2021), autocorrelation in random error was not accounted for, with completely random unfiltered error fields used instead (Wechsler & Kroll, 2006).”
Quote in the paperSubstantial claim1 of 2 sources checked - Interpretationcitation
If multiple cost functions representing the same hypothesis produce similar results, then the shared hypothesis can be deemed robust.
“If multiple cost functions representing the same hypothesis, e.g. the relationship between time taken and slope gradient, each similar but distinct in their idealisation, are able to produce similar results, then the shared hypothesis can be deemed robust (sensu Levins, 1966;Weisberg, 2006;Wimsatt, 2007, pp. 94-132).”
Quote in the paperSubstantial claim1 of 3 sources checked- cites Levins, 1966
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Weisberg, 2006 source
Section 3.2 says that similar but distinct models predicting a common result establish a robust property, which supports the comparison described in the claim. But Weisberg also requires identifying the common structure that generates that result and formulating a conditional theorem; similar results alone do not establish that the shared hypothesis is robust. The Conclusions further distinguish a robust theorem from a confirmed claim about a real phenomenon. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites Wimsatt, 2007 source
- Findingown data
Similar least-cost paths can be produced from slope-based cost functions representing both the same hypothesis and different hypotheses.
“this study demonstrates that similar least-cost paths can be produced from slope-based cost functions representing both the same hypothesis and different hypotheses”
Quote in the paperSubstantial claim - Findingown data
Both model outcomes representing the minimisation of time and energy are unable to explain the placement of the Roman road.
“it is shown that both model outcomes representing the minimisation of time and energy are unable to explain the placement of the Roman road.”
Quote in the paperSubstantial claim
- Findingown data
The two hypotheses produce similar model outcomes when comparing across hypotheses.
“the two hypotheses also produce similar model outcomes when comparing across hypotheses.”
Quote in the paperSubstantial claim - Findingown data
The pattern of small deviation within and across model outcomes is consistent across all landscape complexities and scenarios tested.
“This pattern is consistent across all landscape complexities and scenarios tested.”
Quote in the paperSubstantial claim - Findingnone given
As landscape complexity increases, the slope gradients within the landscapes also increase.
“As landscape complexity increases, the slope gradients within the landscapes also increase (Fig. 2).”
Quote in the paperSubstantial claim - Findingown data
The range within energy-based cost functions increases as landscape complexity increases for both scenarios.
“The range within energy-based cost functions also increases as landscape complexity increases for both scenarios (Fig. 6).”
Quote in the paperNot a substantial claim - Findingown data
A number of least-cost paths representing the hypothesis 'humans minimise energy expenditure when traversing the landscape' follow a route similar to the 'a Karalibus Sulcos' Roman road.
“It should be noted, however, that a number of least-cost paths representing the hypothesis 'humans minimise energy expenditure when traversing the landscape' do follow a route similar to the 'a Karalibus Sulcos' Roman road (Fig. 9B).”
Quote in the paperSubstantial claim - Findingown data
The probabilistic least-cost paths representing the two hypotheses lack similarity to the 'a Karalibus Sulcos' Roman road.
“The lack of similarity between the 'a Karalibus Sulcos' Roman road and the expected outcome given the two hypotheses nonetheless shows that neither the minimisation of time or energy is able to explain the placement of the Roman road, and by extension the decision-making process of the Romans when constructing the road.”
Quote in the paperSubstantial claim - Findingown data
Multiple time-and energy-based cost functions can be used concurrently via probabilistic least-cost paths.
“Despite this epistemic limitation, this study has shown, through case study two, that multiple time-and energy-based cost functions can be used concurrently via probabilistic least-cost paths.”
Quote in the paperSubstantial claim - Findingown data
The two hypotheses remain underdetermined irrespective of distance between the origin and destination used when calculating the LCP.
“The joint increase in both deviation within and across hypotheses as Euclidean distance increases nonetheless indicates that the two hypotheses remain underdetermined irrespective of distance between the origin and destination used when calculating the LCP.”
Quote in the paperSubstantial claim - Findingnone given
Increasing the Euclidean distance between the origin and destination used when calculating the LCP increases the deviation within and across hypotheses.
“Increasing the Euclidean distance between the origin and destination used when calculating the LCP also increases the deviation within and across hypotheses (Fig. 7).”
Quote in the paperNot a substantial claim - Findingown data
For both scenarios, the deviation within and across model outcomes is small.
“Figure 5 shows that for both scenarios, i.e. when elevation is scaled to between 0 to 5 m (Fig. 5A) and 0 to 10 m (Fig. 5B), the deviation within and across model outcomes is small.”
Quote in the paperSubstantial claim - Findingown data
Of the two energy-based cost functions tested, only Herzog (2014c) can produce the model outcomes that follow the Roman road.
“Of the two energy-based cost functions tested, that is Herzog (2014c) and Llobera and Sluckin (2007), it is only however Herzog (2014c) that can produce the model outcomes that follow the Roman road (Fig. 10A).”
Quote in the paperSubstantial claim0 of 2 sources checked - Findingnone given
The hypotheses 'humans minimise time taken when traversing the landscape' and 'humans minimise energy expenditure when traversing the landscape' as represented by multiple time-and energy-based cost functions are robust.
“With the hypotheses 'humans minimise time taken when traversing the landscape' and 'humans minimise energy expenditure when traversing the landscape' as represented by multiple time-and energy-based cost functions shown to be robust, multiple time-and energy-based cost functions were used concurrently.”
Quote in the paperNot a substantial claim - Interpretationreasoning
The Roman road was influenced by pre-existing routes and settlements.
“Rather, it is suggested that the Roman road was influenced by pre-existing routes and settlements.”
Quote in the paperSubstantial claim - Interpretationreasoning
When using LCP analysis to explain known past routes, the hypothesis that multiple cost functions aim to represent, not the cost function itself, is of interest.
“When using LCP analysis to explain known past routes, it is thus not the cost function itself that is of interest-this is merely a simplifying device used to represent the relationship between cost and slope gradient and operationalised to calculate the LCP-but rather the hypothesis that multiple cost functions aim to represent.”
Quote in the paperSubstantial claim - Interpretationreasoning
The use of pre-existing routes and settlements when constructing the Roman road meant that the movement of resources and local traffic across south-western Roman Sardinia could be controlled.
“Irrespective of the exact process, the use of pre-existing routes and settlements when constructing the Roman road meant that the movement of resources and local traffic across south-western Roman Sardinia could be controlled.”
Quote in the paperSubstantial claim - Study design and conductcitation
Testing five different landscape complexities across fractal dimensions 2.20 to 2.60 is deemed to sufficiently capture the variability present in real landscapes.
“With complexity of real landscapes ranging from a fractal dimension of 2.20 to 2.60 (Hofierka et al., 2009), the testing of five different landscape complexities across this range is deemed to sufficiently capture the variability present in real landscapes.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Hofierka et al., 2009
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Premisecitation
DEMs, unless corrected for, can contain modern-day activities that post-date the ancient landscape.
“As discussed by Herzog and Yépez (2015), DEMs-unless corrected for-can contain modern-day activities that post-date the ancient landscape.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Herzog and Yépez (2015)
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationown data
The model outcomes of the hypotheses 'humans minimise time taken / energy expended' as represented by multiple time- and energy-based cost functions are suggested to produce credible realisations of the expected outcome given the hypotheses.
“That is, although the model outcomes of the hypotheses 'humans minimise time taken / energy expended when traversing the landscape' as represented by multiple time-and energy-based cost functions are suggested to produce credible realisations of the expected outcome given the hypotheses, the two hypotheses also produce similar model outcomes when comparing across hypotheses.”
Quote in the paperSubstantial claim - Interpretationreasoning
For LCP model outcomes to credibly explain past decision-making, model outcomes must sufficiently represent the outcome of hypotheses rather than reflecting simplifying assumptions within a single cost function.
“For LCP model outcomes to credibly explain the decision-making processes of past people when traversing the landscape, it is therefore necessary that model outcomes sufficiently represent the outcome of hypotheses, e.g.time or energy, rather than reflecting the specific simplifying assumptions within a single cost function”
Quote in the paperNot a substantial claim - Interpretationreasoning
The model outcomes following the Roman road are more likely to be the product of model idealisation and not representative of the hypothesis but artefacts of the specific cost function.
“Given this, it is suggested that these model outcomes are more likely to be the product of model idealisation, and thus are not representative of the hypothesis but the artefacts of the specific cost function.”
Quote in the paperSubstantial claim - Interpretationown data
The underdetermination impacts the support for choosing which hypothesis best explains the known past route when using LCP analysis, and by extension the explanatory power of these hypotheses for understanding past decision-making.
“This underdetermination impacts the support for choosing which hypothesis best explains the known past route when using LCP analysis, and by extension the explanatory power of these hypotheses for understanding past decision-making.”
Quote in the paperSubstantial claim - Interpretationreasoning
Results are proposed to be applicable when calculating LCPs using DEMs with increasingly higher resolution from the same origin and destination.
“Given that increasing the Euclidean distance also increases the number of cells between the origin and destination, these results are also proposed to be applicable when calculating LCPs using DEMs with increasingly higher resolution from the same origin and destination.”
Quote in the paperNot a substantial claim - Interpretationreasoning
The increasing range within energy-based cost functions as landscape complexity increases can be attributed to the process of model idealisation.
“The increasing range as landscape complexity increases within the energy-based cost functions can be attributed to the process of model idealisation.”
Quote in the paperSubstantial claim - Interpretationreasoning
Similar model outcomes can be produced irrespective of whether the minimisation of time or energy hypothesis is employed.
“Given this, similar model outcomes can be produced irrespective of whether the minimisation of time or energy hypothesis is employed.”
Quote in the paperSubstantial claim - Interpretationreasoning
As slope gradient increases, differences in cost functions become more pronounced, resulting in a greater impact on the model outcome.
“As slope gradient increases, differences in cost functions become more pronounced, resulting in a greater impact on the model outcome.”
Quote in the paperSubstantial claim - Premisecitation
No single cost function can simultaneously maximise accuracy, precision, generality, and simplicity.
“As a result, no single cost function can simultaneously maximise all these properties (Levins, 1966).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Levins, 1966
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Premisecitation
The critical gradient parameter within the wheeled vehicle cost function developed by Llobera and Sluckin (2007) is not given a predefined value defined by the cost function but instead to be estimated from the past route being explained.
“Although the hypothesis could be formulated as 'humans minimise steeper slope gradients when traveling by wheeled vehicles', the critical gradient parameter within the cost function is not given a predefined value defined by the cost function but instead to be estimated from the past route being explained.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Llobera and Sluckin (2007) source
- Premisereasoning
Cost functions contain idealisations; they intentionally misrepresent the relationship between an associated cost and slope gradient through different assumptions, approximations, and simplifications.
“Cost functions, themselves models, also contain idealisations; that is, they intentionally misrepresent-through different assumptions, approximations, and simplifications-the relationship between an associated cost and slope gradient.”
Quote in the paperSubstantial claim - Interpretationreasoning
Model outcomes from multiple cost functions representing the same hypothesis are expected to slightly differ because each cost function is its own idealisation.
“The need for multiple cost functions to produce similar-but not identical-model outcomes reflects that each cost function is its own idealisation.As a result, model outcomes are expected to slightly differ.”
Quote in the paperSubstantial claim - Interpretationreasoning
The use of pre-existing routes and settlements played a greater role in the placement of the 'a Karalibus Sulcos' Roman road than minimising time or energy when traversing on foot from the western coast to the modern-day town of Decimomannu.
“More generally, the lack of similarity between the probabilistic least-cost paths and the Roman road suggests that the use of pre-existing routes and settlements therein played a greater role in its placement than minimising time or energy when traversing on foot from the western coast to the modern-day town of Decimomannu.”
Quote in the paperSubstantial claim - Interpretationreasoning
If robustness is not present, model outcomes can be deemed to not sufficiently represent the hypothesis of interest and thus do not provide a credible realisation of the expected outcome given the hypothesis.
“Conversely, if robustness is not present, model outcomes can be deemed to not sufficiently represent the hypothesis of interest and thus do not provide a credible realisation of the expected outcome given the hypothesis.”
Quote in the paperNot a substantial claim - Interpretationreasoning
It is difficult to discern which hypothesis produced the model outcome, limiting the explanatory power of these hypotheses for understanding the decision-making process used by people in the past.
“As a result, it is difficult to discern which hypothesis produced the model outcome, thereby impacting the support for choosing which hypothesis best explains the known past route, and by extension limits the explanatory power of these hypotheses for understanding the decision-making process used by people in the past.”
Quote in the paperSubstantial claim - Interpretationreasoning
The underdetermination of the two hypotheses shows that it is more likely that the model outcomes representing the two hypotheses will not be distinguishable in their realisation.
“The underdetermination of the two hypotheses does however show that it is more likely that the model outcomes representing the two hypotheses will not be distinguishable in their realisation”
Quote in the paperSubstantial claim - Interpretationcitation
The effect of the random error fields can be deemed as the worst-case scenario.
“As a result of the second advantage, the effect of the random error fields can be deemed as the worst-case scenario (Wechsler & Kroll, 2006).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Wechsler & Kroll, 2006 source
- Interpretationreasoning
Any modelled outcome from a cost function represents the artefacts of the specific mathematical formalisation used to represent the hypothesis and not the hypothesis itself.
“Thus, any modelled outcome from this cost function would represent the artefacts of the specific mathematical formalisation used to represent the hypothesis and not the hypothesis itself.”
Quote in the paperNot a substantial claim - Interpretationreasoning
When multiple cost functions representing the same hypothesis are used concurrently, the impact of the process of model idealisation on the model outcome is reduced.
“Thus, when multiple cost functions representing the same hypothesis are used concurrently, the impact of the process of model idealisation on the model outcome is reduced.”
Quote in the paperSubstantial claim - Interpretationreasoning
Through the use of multiple time- and energy-based cost functions and the incorporation of random error in the DEM, spurious results in calculated LCPs as a result of model idealisation and measurement error in the DEM will be reduced.
“Through the use of multiple timeand energy-based cost functions and the incorporation of random error in the DEM, spurious results in calculated least-cost paths as a result of model idealisation when creating the cost functions and measurement error in the DEM will be reduced.”
Quote in the paperSubstantial claim - Interpretationreasoning
Results from the tactical simulation will be applicable across multiple landscape complexities and LCP analysis studies.
“With this, results presented here will be applicable across multiple landscape complexities and LCP analysis studies.”
Quote in the paperSubstantial claim - Interpretationnone given
If model outcomes using cost functions with different hypotheses are indistinguishable in their realisations, then the two hypotheses are underdetermined and do not provide adequate support for choosing which hypothesis best explains past decision-making.
“Or worse, if model outcomes using cost functions with different hypotheses are indistinguishable in their realisations, then the two hypotheses are underdetermined and thus do not provide adequate support for choosing which hypothesis best explains the decision-making processes used by past people when traversing the landscape.”
Quote in the paperSubstantial claim - Study design and conductcitation
The topography-only simulations provide a 'laboratory' where all other factors are not present and thus cannot influence the modelled results.
“Rather, the topography-only simulations provide a 'laboratory' where all other factors are not present and thus cannot influence the modelled results (sensu Bevan, 2013).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Bevan, 2013
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
A cost function, even with a pre-defined critical gradient value, is only a single model.
“More importantly, the cost function itself, even with a pre-defined critical gradient value, is only a single model.”
Quote in the paperNot a substantial claim - Interpretationcitation
By producing model outcomes that sufficiently represent the outcome of the hypothesis and not artefacts of a single cost function, model outcomes can more credibly be used for explaining known past routes and past decision-making.
“By producing model outcomes that sufficiently represent the outcome of the hypothesis and not the artefacts of a single cost function (sensu Levins, 1966), i.e. the hypothesis is shown to be robust, model outcomes can more credibly be used for explaining known past routes and more importantly the decision-making processes of past people.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Levins, 1966
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
The robustness of the two hypotheses suggests that the concurrent use of multiple time- and energy-based cost functions sufficiently represents the hypotheses 'humans minimise time taken / energy expenditure', respectively.
“Despite the issue of underdetermination, the robustness of the two hypotheses does however suggest that the concurrent use of multiple time-and energy-based cost functions sufficiently represents the hypotheses 'humans minimise time taken / energy expenditure when traversing the landscape', respectively.”
Quote in the paperSubstantial claim - Interpretationreasoning
Only when model outcomes sufficiently represent the outcome of the hypothesis of interest (i.e., the hypothesis is robust) can LCP model outcomes credibly be used for explaining why people in the past moved where they did.
“Only when model outcomes sufficiently represent the outcome of the hypothesis of interest, i.e. the hypothesis is robust, can LCP model outcomes credibly be used for explaining why people in the past moved where they did.”
Quote in the paperSubstantial claim
- Probably not a substantial claim. 8
- It may state two separate claims as one. 13
Probably not a substantial claim.
PremisecitationAll scientific models contain idealisations: the intentional misrepresentation of a target system.
“All scientific models contain idealisations: the intentional misrepresentation of a target system.”
Quote in the paperSubstantial claim1 of 3 sources checked- cites McMullin, 1985 source
- cites Potochnik, 2017 source
- cites Weisberg, 2007 source
In “Additional information” (page unavailable), Weisberg says that all three forms of idealization involve the willful distortion of scientific representations, supporting the definition of idealization as intentional misrepresentation. The passage does not say that every scientific model contains an idealization; “all three forms” refers to the forms of idealization, not to all models. (LLM, conf=0.88)
Experimental — accuracy not yet measured
Probably not a substantial claim.
Study design and conductcitationThe uncertainty in the LCPs as a result of random error in the DEM was propagated through the analysis, with each LCP calculated fifty times with different realisations of random error.
“The uncertainty in the LCPs as a result of random error in the DEM was propagated through the analysis, with each LCP calculated fifty times with different realisations of random error (see Lewis, 2021 for methodology).”
Quote in the paperSubstantial claim1 of 1 sources checked- cites Lewis, 2021 source
It may state two separate claims as one.
Findingown dataLeast-cost path results are robust but underdetermined under the tested environmental settings.
“suggesting that least-cost path results are robust but underdetermined under the tested environmental settings.”
- Also stated in: § Results and Discussion · plain
“whilst the two hypotheses are robust, they are also underdetermined (Fig. 4C outcome).”
- Also stated in: § Results and Discussion · plain
“the model outcomes from both hypotheses 'humans minimise time taken / energy expenditure when traversing the landscape' as represented by multiple time-and energy-based cost functions are robust but underdetermined across landscapes with increasing complexity, different maximum slope gradients, and increasing distances between origin and destination locations when calculating the least-cost path.”
Quote in the paperSubstantial claim- Also stated in: § Results and Discussion · plain
It may state two separate claims as one.
Findingown dataIt is difficult to discern between model outcomes from the two hypotheses; the expected deviation within the two hypotheses coincides with the expected deviation across hypotheses.
“With this, it is difficult to discern between model outcomes from the two hypotheses; the expected deviation within the two hypotheses coincides with the expected deviation across hypotheses.”
Quote in the paperSubstantial claimProbably not a substantial claim.
PremisecitationLeast-cost path analysis is predicated on the assumption that humans, whether modern or past, use all available information to economise their behaviour when traversing the landscape.
“LCP analysis is predicated on the assumption that humans-whether modern or past-use all available information to economise their behaviour when traversing the landscape (Surface-Evans & White, 2012;Zipf, 1949).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Surface-Evans & White, 2012 source
- cites Zipf, 1949
Probably not a substantial claim. The paper may give no reason why it holds.
PremisereasoningWhen an LCP calculated using a time-based cost function representing 'humans minimised time taken' shares similarity with a known past route, it can be inferred that the placement of the known past route was chosen to minimise time.
“For example, if an LCP calculated using a time-based cost function representing the hypothesis 'humans minimised time taken when traversing the landscape' shares similarity with a known past route, it can be inferred that the placement of the known past route was chosen to minimise time.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
PremisecitationGüimil-Fariña and Parcero-Oubiña (2015) compared four cost functions, despite three aiming to represent the same hypothesis 'humans minimise energy expenditure when traversing the landscape'.
“For example, Güimil-Fariña and Parcero-Oubiña (2015) compared four cost functions, despite three of these aiming to represent the same hypothesis 'humans minimise energy expenditure when traversing the landscape'.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Güimil-Fariña and Parcero-Oubiña (2015) source
It may state two separate claims as one.
InterpretationreasoningLeast-cost paths calculated using single cost functions are liable to spurious results with limited power for explaining known past routes and past decision-making.
“As a result, least-cost paths calculated using single cost functions are liable to spurious results with limited power for explaining known past routes, and by extension the decision-making processes of past people.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningCurrent approaches to least-cost path analysis misrepresent what cost functions are, their relationship to the hypotheses they aim to represent, and their role in explanation.
“Despite their prevalence, current approaches misrepresent what these cost functions are, their relationship to the hypotheses that they aim to represent, and their role in explanation.”
- Also stated in: § Introduction · plain
“this approach misrepresents what cost functions are, their relationship to the hypotheses that they aim to represent, and their role in explanation.”
Quote in the paperSubstantial claim- Also stated in: § Introduction · plain
It may state two separate claims as one.
PremisecitationWithin multiple model idealisation, each individual model is deemed false but might still be useful when combined.
“Within MMI, each individual model is deemed to be false but might still be useful when combined (Wimsatt, 1981(Wimsatt, , 1987)).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Wimsatt, 1981
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites Wimsatt, 1987
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
InterpretationreasoningAccounting for model idealisation and measurement error when using least-cost path analysis strengthens the explanatory power of the two hypotheses and reduces the risk of spurious results and interpretations.
“Although limited by the issue of underdetermination, by accounting for the process of model idealisation and measurement error when using least-cost path analysis, the explanatory power of the two hypotheses is strengthened, with the risk of spurious results and interpretations reduced.”
Quote in the paperSubstantial claimProbably not a substantial claim.
InterpretationreasoningAny modelled outcome that aims to represent the hypothesis 'humans minimise steeper slope gradients than x critical gradient when traveling by wheeled vehicles' should include an additional clause.
“Therefore, any modelled outcome that aims to represent the hypothesis should include an additional clause: 'humans minimise steeper slope gradients than x critical gradient when traveling by”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningDifferences in model outcomes from cost functions sharing the same hypothesis can be examined, with the reason for these differences better understood.
“Furthermore, differences in model outcomes from cost functions sharing the same hypothesis can be examined, with the reason for these differences better understood.”
Quote in the paperSubstantial claimProbably not a substantial claim.
PremisecitationThe 'a Karalibus Sulcos' Roman road provided an internal and more direct route than the road along the coast via Teluga and Nura.
“Providing an internal and more direct route than the road along the coast via Teluga and Nura (Fig. 8), the road was used to transport lead silver and wheat stored at Sulci to the port at Carales before being distributed to Italy (Atzori, 2006, pp. 11-13).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Atzori, 2006
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
InterpretationreasoningEach hypothesis becomes independent of one another and is not shared across cost functions with different critical gradient values.
“As a result of this, each hypothesis become independent of one another and is not shared across cost functions with different critical gradient values.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningUsing probabilistic least-cost paths reduces the impact of model idealisation on the model outcome and accounts for random error in the digital elevation model.
“This both reduces the impact of model idealisation on the model outcome, whilst also accounting for random error in the digital elevation model.”
Quote in the paperSubstantial claimProbably not a substantial claim.
PremisecitationThe 'a Karalibus Sulcos' Roman road was used to transport lead silver and wheat stored at Sulci to the port at Carales before being distributed to Italy.
“the road was used to transport lead silver and wheat stored at Sulci to the port at Carales before being distributed to Italy (Atzori, 2006, pp. 11-13).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Atzori, 2006
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
It may state two separate claims as one.
InterpretationcitationThe use of unfiltered error fields reduces computational burden and makes no assumptions about the spatial relationships of the random error.
“The use of unfiltered error fields has two advantages: (1) reduction in computational burden as spatial autocorrelation does not need to be calculated, and (2) no assumptions are made about the spatial relationships of the random error.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Interpretationown dataNeither the minimisation of time nor energy is able to explain the placement of the 'a Karalibus Sulcos' Roman road.
“Thus, the process of road construction for the 'a Karalibus Sulcos' Roman road cannot be credibly attributed to the desire to minimise time or energy when traversing on foot from the western coast to the modern-day town of Decimomannu, or, in short, these hypotheses are rejected as explaining the Roman road.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningThe probabilistic least-cost path aims to reflect the expected outcome given the hypothesis and not the artefacts of a single model.
“With this, the probabilistic least-cost path aims to reflect the expected outcome given the hypothesis and not the artefacts of a single model.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
PremisecitationWith LCP results previously shown to be impacted by random error within the DEM, probabilistic LCPs were used within the Roman road case study.
“With least-cost path analysis results previously shown to be impacted by random error within the DEM (Lewis, 2021), probabilistic least-cost paths were used within this case study.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Lewis, 2021 source
Checked before — its earlier verdict is held back until a new check runs.
The paper relies on this without support.
The 'a Karalibus Sulcos' Roman road was constructed for traversal on foot.
“when traversing on foot from the western coast to the modern-day town of Decimomannu”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The way travel costs are quantified and mapped is a problem common to all forms of digital mobility modelling and has been discussed at length in archaeology.
“This is a problem common to all forms of digital mobility modelling and has been discussed at length, in particular in archaeology (Herzog, 2013b;Kantner, 2012;Lewis, 2024).”
From Exploring the inherent affordability for movement in past landscapes (2027) · cites this paper as Lewis, 2024
The Abstract and Introduction discuss problems with how walking costs are represented by slope-based cost functions in archaeological least-cost-path modelling. They support the claim that this issue has been discussed in archaeology, but do not establish that it is common to all forms of digital mobility modelling. Page numbers are not available in our copy. (LLM, conf=0.86)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Pairwise distances counted as least-cost paths
MinorThe 91 values per simulation are pairwise distances between 14 least-cost paths, not 91 paths. Across five landscapes and 1000 simulations, that gives 70,000 least-cost paths per scenario, not 455,000.
Reported 455000 · recomputed 70000
“The approach outlined above was repeated 1000 times, generating two sets of 455,000 least-cost paths (91 LCPs * 5 landscapes * 1000 simulations for each scenario)”
Citations
Cites
Re-Engineering Philosophy for Limited Beings: Piecewise Approximations to Reality
WIMSATT, W. C. (2007) · doi:10.2307/j.ctv1pncnrh · Cited by 871
Weisberg, M. (2013) · Oxford University Press eBooks · doi:10.1093/acprof:oso/9780199933662.001.0001 · Cited by 772
Energy cost of walking and running at extreme uphill and downhill slopes
Minetti, A. E., Moia, C., Roi, G. S., Susta, D., & Ferretti, G. (2002) · Journal of Applied Physiology · doi:10.1152/japplphysiol.01177.2001 · Cited by 624
Book Review: Scientific Inquiry and the Social Sciences
Cox, E. P., Brewer, M. B., & Collins, B. E. (1983) · Journal of Marketing Research · doi:10.2307/3151691 · Cited by 543
Algorithms for random fractals
Saupe, D. (1988) · Springer eBooks · doi:10.1007/978-1-4612-3784-6_2 · Cited by 510
Predicting energy expenditure with loads while standing or walking very slowly
Pandolf, K. B., Givoni, B., & Goldman, R. F. (1977) · Journal of Applied Physiology · doi:10.1152/jappl.1977.43.4.577 · Cited by 474
Idealization and the Aims of Science
Potochnik, A. (2017) · doi:10.7208/chicago/9780226507194.001.0001 · Cited by 421
Weisberg, M., & Journal of Philosophy, I. (2007) · Journal of Philosophy · doi:10.5840/jphil20071041240 · Cited by 418
McMullin, E. (1985) · Studies in History and Philosophy of Science Part A · doi:10.1016/0039-3681(85)90003-2 · Cited by 387
The Explanation of Culture Change
(2017) · Oxbow Books · doi:10.2307/j.ctt1vgwb5m.11 · Cited by 378
A simple positional accuracy measure for linear features
Goodchild, M. F., & Hunter, G. J. (1997) · International Journal of Geographical Information Systems · doi:10.1080/136588197242419 · Cited by 337
Nitecki, M. H., & Hoffman, A. (1987) · Cited by 328
Barrington Atlas of the Greek and Roman World
Johnson, J. M. (2000) · Cartographic Perspectives · doi:10.14714/cp37.816 · Cited by 326
Weisberg, M. (2006) · Philosophy of Science · doi:10.1086/518628 · Cited by 223
False Models as Means to Truer Theories
(2007) · Harvard University Press eBooks · doi:10.2307/j.ctv1pncnrh.10 · Cited by 215
Robustness, Reliability, and Overdetermination (1981)
Wimsatt, W. C. (2012) · Boston studies in the philosophy of science · doi:10.1007/978-94-007-2759-5_2 · Cited by 178
THREE PRESENTATIONS ON GEOGRAPHICAL ANALYSIS AND MODELING
Modeling, N.-I., & Tobler, W. (1993) · Cited by 158
Least-cost paths in mountainous terrain
Rees, W. G. (2004) · Computers & Geosciences · doi:10.1016/j.cageo.2003.11.001 · Cited by 156
Quantifying DEM Uncertainty and its Effect on Topographic Parameters
Wechsler, S. P., & Kroll, C. N. (2006) · Photogrammetric Engineering & Remote Sensing · doi:10.14358/pers.72.9.1081 · Cited by 148
The Quality of the Archaeological Record
Perreault, C. (2018) · University of Chicago Press eBooks · doi:10.7208/chicago/9780226631011.001.0001 · Cited by 135
Zigzagging: Theoretical insights on climbing strategies
Llobera, M., & Sluckin, T. J. (2007) · Journal of Theoretical Biology · doi:10.1016/j.jtbi.2007.07.020 · Cited by 103
Least-cost Paths – Some Methodological Issues
Herzog, I. (2014) · Internet Archaeology · doi:10.11141/ia.36.5 · Cited by 102
Güimil-Fariña, A., & Parcero-Oubiña, C. (2015) · Journal of Archaeological Science · doi:10.1016/j.jas.2014.11.030 · Cited by 95
Using crowdsourced fitness tracker data to model the relationship between slope and travel rates
Campbell, M. J., Dennison, P. E., Butler, B. W., & Page, W. G. (2019) · Applied Geography · doi:10.1016/j.apgeog.2019.03.008 · Cited by 87
Barreca, F. (1979) · Cited by 74
Modeling Métis mobility? Evaluating least cost paths and indigenous landscapes in the Canadian west
Supernant, K. (2017) · Journal of Archaeological Science · doi:10.1016/j.jas.2017.05.006 · Cited by 64
Measuring and modeling the speed of human navigation
Irmischer, I. J., & Clarke, K. C. (2018) · Cartography and Geographic Information Science · doi:10.1080/15230406.2017.1292150 · Cited by 62
Estimated travel time for walking trails in natural areas
Márquez-Pérez, J., Vallejo-Villalta, I., & Álvarez-Francoso, J. I. (2017) · Geografisk Tidsskrift-Danish Journal of Geography · doi:10.1080/00167223.2017.1316212 · Cited by 56
Lewis, J. (2021) · Journal of Archaeological Method and Theory · doi:10.1007/s10816-021-09522-w · Cited by 53
Self-citationChapter 17 Geomorphometry in GRASS GIS
Hofierka, J., Mitášová, H., & Neteler, M. (2008) · Developments in psychiatry · doi:10.1016/s0166-2481(08)00017-2 · Cited by 48
TINITALY, a digital elevation model of Italy with a 10 meters cell size, version 1.0
Tarquini, S., Isola, I., Favalli, M., & Battistini, A. (2020) · Istituto Nazionale di Geofisica e Vulcanologia (INGV) · doi:10.13127/tinitaly/1.0 · Cited by 46
Spatial analysis based on cost functions
Herzog, I. (2020) · Archaeological Spatial Analysis · doi:10.4324/9781351243858-18 · Cited by 45
Fonte, J., Parcero‐Oubiña, C., & García, J. M. C. (2017) · Mediterranean Archaeology & Archaeometry. International Journal/Mediterranean archaeology and archaeometry. International scientific journal · doi:10.5281/zenodo.1005561 · Cited by 43
Least Cost Analysis of Social Landscapes: Archaeological Case Studies
(2012) · doi:10.1353/book41407 · Cited by 41
1. Making Human Space: The Archaeology of Trails, Paths, and Roads
Snead, J. E., Erickson, C. L., & Darling, J. A. (2010) · University of Pennsylvania Museum of Archaeology and Anthropology eBooks · doi:10.2307/j.ctt3fhjb3.7 · Cited by 36
Mastino, A. (2005) · Cited by 30
Issues in replication and stability of least-cost path calculations
Herzog, I. (2022) · Studies in DIGITAL Heritage · doi:10.14434/sdh.v5i2.33796 · Cited by 23
Parcero-Oubiña, C., Güimil-Fariña, A., Fonte, J., & Costa-García, J. M. (2019) · Computational Social Sciences · doi:10.1007/978-3-030-04576-0_14 · Cited by 20
A least cost analysis: correlative modeling of the chaco regional road system
Field, S., Heitman, C., & Richards-Rissetto, H. (2019) · Journal of Computer Applications in Archaeology · doi:10.5334/jcaa.36 · Cited by 18
Pace and Critical Gradient for Hill Runners: An Analysis of Race Records
Kay, A. (2012) · Journal of Quantitative Analysis in Sports · doi:10.1515/1559-0410.1456 · Cited by 16
Review of Least Cost Analysis of Social Landscapes. Archaeological Case Studies [Book]
Herzog, I. (2013) · Internet Archaeology · doi:10.11141/ia.34.7 · Cited by 12
Arca, G. A. (2017) · DOAJ (DOAJ: Directory of Open Access Journals) · doi:10.13125/2532-0289/3100 · Cited by 4
Running Uphill: An Experimental Result and Its Applications
Davey, R. C., Hayes, M., & Norman, J. M. (1994) · Journal of the Operational Research Society · doi:10.2307/2583947 · Cited by 3
Testing Models for Medieval Settlement Location
Herzog, I. (2014) · Studies in Classification, Data Analysis, and Knowledge Organization · doi:10.1007/978-3-319-01595-8_38 · Cited by 2
Traveling in a Prehistoric Landscape: Exploring the Influences that Shaped Human Movement
Unknown, (. (2010) · Universitätsbibliothek Tübingen · doi:10.15496/publikation-3314
(1949) · Max Planck Digital Library
Unknown, (. (2010) · Universitätsbibliothek Tübingen · doi:10.15496/publikation-3307
(2013) · Dialnet (Universidad de la Rioja)
Fonte, J., Parcero-Oubiña, C., & Costa-García, J. M. (2017) · Zenodo · doi:10.5281/zenodo.1005562
Flagged first, then by citations
3 of the 52 references couldn’t be matched to a known paper.
Research works printed but not linked: 11
- Atzori, S. (2006). La strada romana “a Karalibus Sulcos.” Mogoro: PTM.
- Bevan, A. (2013). Travel and interaction in the Greek and Roman world. A review of some computational modelling approaches. Bulletin of the Institute of Classical Studies. Supplement, 122, 3–24.
- Garmy, P., Kaddouri, L., Rozenblat, C., & Schneider, L. (2005). Logiques spatiales et “systèmes de villes” en Lodévois de l’Antiquité à la période moderne. In Temps et espaces de l’homme en société, analyses et modèles spatiaux en archéologie (pp. 335–346).
- Herzog, I., & Yépez, A. (2015). The impact of the DEM on archaeological GIS studies: A case study in Ecudaor. Presented at the Proceedings of the 19th International Conference on Cultural Heritage and New Technologies 2014 (CHNT 19, 2014)., Vienna. https://www.chnt.at/wp-content/uploads/eBook_CHNT20_Herzog_Yepez_2015.pdf. Accessed 11 Sept 2023.
- Levins, R. (1966). The strategy of model building in population biology. American Scientist, 54(4), 421–431.
- Meloni, P. (1990). La Sardegna Romana (2nd ed.). Sassari: Chiarella.
- Naismith, W. (1892). Excursions: Cruach Ardran, Stobinian, and Ben More. Scottish Mountaineering Club Journal, 2(3), 136.
- Pietra, G. (2015). Il Sulcis in età romana. L’Africa romana: Momenti di continuità e rottura: Bilancio di trent’anni di convegni, Atti del XX Convegno Internazionale di studi (Alghero - Porto Conte Ricerche, 26–29 settembre 2013) (Vol. III, pp. 1913–1920). Carocci.
- Tate, N. J., & Wood, J. (2001). Fractals and scale dependencies in topography. In N. J. Tate & P. Atkinson (Eds.), Scale in Geographical Information Systems (pp. 35–51). New York: John Wiley & Sons.
- Tetti, V. (1985). Antiche vie romane della Sardegna e cursus publicus: note e riferimenti toponomastico. Archivio Storico Sardo di Sassari, XI, 79–119.
- Tronchetti, C. (1995). Le problematiche del territorio del Sulcis in età romana. In V. Santoni (Ed.), Carbonia e il Sulcis. Archeologia e territorio (pp. 265–275). S’Alvure: Oristano.
Cited by
Sorted by citations
Saintenoy, T., Llobera, M., Thiéry, N. M., Crespo Fernández, M., Fábrega-Álvarez, P., & Santos, R. (2025) · Journal of Archaeological Science · doi:10.1016/j.jas.2024.106125 · Cited by 9
Álvarez, P. A. A., & Fernández, P. T. (2025) · Archaeological and Anthropological Sciences · doi:10.1007/s12520-025-02175-w · Cited by 7
Hagmann, D. (2025) · Journal of Archaeological Science · doi:10.1016/j.jas.2024.106151 · Cited by 6
A GIS Analysis of Coastal Proximity with a Prehistoric Greek Case Study
Nuttall, C. (2024) · Journal of Computer Applications in Archaeology · doi:10.5334/jcaa.143 · Cited by 5
Resource dependency and communication networks in Early Neolithic western Europe
Kempf, M., & Denis, S. (2024) · Quaternary Environments and Humans · doi:10.1016/j.qeh.2024.100014 · Cited by 5
Daems, D., & Kafetzaki, D. (2024) · Journal of Archaeological Science · doi:10.1016/j.jas.2024.106121 · Cited by 4
Filling the Gaps—Computational Approaches to Incomplete Archaeological Networks
Priß, D., Wainwright, J., Lawrence, D., Turnbull, L., Prell, C., Karittevlis, C., & Ioannides, A. A. (2025) · Journal of Archaeological Method and Theory · doi:10.1007/s10816-024-09688-z · Cited by 4
Frozen motion: Contextualizing wheel rut data within and beyond the Pompeiian street grid
Picker-Kille, D. (2025) · Journal of Archaeological Science · doi:10.1016/j.jas.2025.106155 · Cited by 2
Least cost path analysis and fluvial navigation of Patagonian hunter-gatherers
Scartascini, F. L., & Gavuzzo, A. B. (2025) · Journal of Archaeological Science · doi:10.1016/j.jas.2025.106397 · Cited by 1
Sunseri, J. U., Castillo, M. A., & Lin, E. (2026) · Kiva · doi:10.1080/00231940.2026.2646017
Exploring the inherent affordability for movement in past landscapes
Parcero-Oubiña, C., & Solana-Muñoz, A. (2027) · Journal of Archaeological Method and Theory · doi:10.1007/s10816-026-09813-0
Fernández, P. T., & Álvarez, P. A. A. (2025) · Acta Archaeologica · doi:10.1163/16000390-09501016
Pažout, A. (2026) · Journal of Archaeological Method and Theory · doi:10.1007/s10816-026-09766-4
A data-driven framework for inferring past movement decisions from archaeological routesPreprint
Lewis, J. (2026) · Research Square · doi:10.21203/rs.3.rs-8948011/v1
Self-citationLewis, J., & Crema, E. R. (2026) · Humanities and Social Sciences Communications · doi:10.1057/s41599-026-09179-4
Self-citation
Metadata
- Authors
- Lewis, J.
- Affiliations
- University of Cambridge (GB)
- Corresponding author
- Lewis, Joseph
- ORCID
- Lewis, Joseph
- Venue
- Journal of Archaeological Method and Theory
- Published
- 2023-09-15
- Volume
- 31
- Issue
- 3
- Pages
- 854-874
- ISSN
- 1072-5369 (ISSN-L)
- 1573-7764
- OpenAlex
- W4386727108
- Semantic Scholar
- a66f239a17a3130f188f586cdd9cd7cbebd8205a
- Cited by
- 15
- References
- 52
- Citations by year
- 2026: 5 · 2025: 6 · 2024: 4
- FWCI
- 8.17
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Journal of Archaeological Method and Theory (publishedVersion, cc-by)
- Apollo (University of Cambridge) (submittedVersion, cc-by)
- Topics
- Social Sciences › Archeology › Archaeological Research and Protection
- Social Sciences › Urban Studies › Urban Design and Spatial Analysis
- Social Sciences › Arts and Humanities › Archeology (arts and humanities) › Archaeological and Geological Studies
- Keywords
- least-cost path analysis, cost function, underdetermination, Roman roads, Sardinia, tactical simulation, robustness analysis, digital elevation model error, settlement location, human mobility
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-08, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-08 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
