Pyrite Framboid Sizes and Marine Redox Conditions

Pyrite framboids are tiny, roughly spherical aggregates of pyrite microcrystals that can preserve clues about marine sedimentary conditions. Yet their environmental meaning depends on entire size distributions, mineral textures, formation pathways and preservation. A single small grain cannot tell us whether ancient seawater was euxinic, and a single large grain cannot certify an oxygenated sea.

Did you know that pyrite can form inside reducing sediment even when the overlying water contains oxygen? Framboids may grow in sediment pores or within sulfide-bearing water, and these different pathways can create contrasting size populations. The art of reading them is to connect measurement geometry and statistics to a plausible formation history without overlooking reworking or later growth.

This guide explains framboid microscopy, diameter conversion, sample statistics and interpretations from first principles. It includes constructed worked examples, independent problems with answers, errors and repairs, and primary methodological sources. Every numerical teaching sample is invented, not represented as a particular field observation.

The first question is what a framboid is

A pyrite framboid is a cluster of very small iron-disulfide crystals arranged into a roughly spherical or raspberry-like aggregate. Its overall diameter may be only a few micrometres, yet it can contain many smaller crystallites. A microscope image can show the overall shape, internal crystal texture and later overgrowths. The mineral’s chemical formula is FeS2, but morphology provides evidence about how crystals nucleated, aggregated and continued growing, not merely the final mass of pyrite.

Framboid formation depends on reactive iron, sulfide availability, supersaturation, nucleation kinetics and the time during which a particle remains in a growth-favourable environment. A particle forming in a sulfide-bearing water column may spend a limited time suspended before sinking. Another growing within sediment pore water may have access to reactants over a longer or different time. These mechanisms can generate different size distributions, although sedimentary conditions and later alteration can blur the distinction.

The original work comparing modern environments found that small, relatively uniform framboids were characteristic of certain modern euxinic settings, whereas deposits formed beneath more oxygenated waters often had larger and more variable populations. The comparison is empirical and mechanistically motivated, not a universal size-to-oxygen conversion equation. It works best when primary framboids are preserved and competing formation pathways have been investigated.

A framboid’s presence does not prove anoxia in the overlying bottom water. Sulfate reduction can generate sulfide below the sediment-water interface even if the overlying water contains oxygen. Pyrite can then precipitate within those pores and leave a microscopic record. The interpreter must distinguish the environment of formation from the environment of burial, because an aggregate can move between them before becoming part of a rock.

Water-column euxinia means the presence of free dissolved sulfide in an anoxic body of water. Anoxia means the absence of oxygen but does not automatically establish dissolved sulfide. Dysoxia refers to depleted oxygen, although numerical definitions vary among studies. These are environmental descriptions, not mineral names. A histogram may support a comparison with known settings but cannot reconstruct a precise ancient oxygen concentration or sulfide concentration without independent evidence.

An ancient core includes layers deposited over time, and one grain may be much older than the enclosing layer if reworking occurred. A turbidite can move grains downslope; a storm can resuspend shallow material; burrowers can transfer particles across depths. Even a size population that formed under sulfide-rich conditions might be redeposited later. The difference between where a grain formed and where it was finally buried should be explicit in the results.

How to recognise primary and altered textures

Scanning electron microscopy can resolve framboid outlines and the constituent microcrystals in polished or broken surfaces. Optical microscopy and reflected-light petrography may also show larger aggregates and relationships with surrounding minerals. Each method has a different detection limit, contrast mechanism and preparation bias. A reported size distribution must specify the imaging approach rather than assuming every instrument sees the same population.

The overall framboid and each microcrystal have separate sizes. Measuring individual microcrystals instead of the entire aggregate will generate a wholly different dataset. Some studies use both: the aggregate diameter helps characterize framboid population structure, while microcrystal size and overgrowth textures help distinguish early formation from recrystallisation. A methods section should identify exactly which boundary was traced.

A late euhedral pyrite crystal is not a framboid simply because it is adjacent to a framboidal cluster. Individual crystals may grow over original aggregates, obscuring their edges and making the apparent diameter larger. Distinguish intact primary framboids, overgrown framboids, loose microcrystal clusters and unrelated massive pyrite. If an observer merges them into one measurement class, a useful environmental distribution may become meaningless.

Some aggregates are packed inside fossil cavities, burrow fills, cracks or later veins. These textural contexts may identify later precipitation, concentration of reactants or localized redox conditions that differ from the original depositional environment. A framboid in a fracture cannot automatically be used to reconstruct conditions at the time of deposition. Describe its host fabric and cross-cutting relationships before including it in the primary dataset.

Oxidation after exposure can destroy or change pyrite. Iron oxides and sulfates may form near weathered surfaces, while partial replacement can alter edges or leave pseudomorphs. Samples with strong weathering may preserve some shape information but lose others. An analyst should document whether measurements come from fresh interior rock or exposed altered material, and whether secondary replacement is evident in the images.

Metamorphism, pressure solution and recrystallisation can modify grain shape and mineral association. Some ancient deposits still preserve framboids despite substantial geological age, but age alone is not proof of preservation. The correct question is whether the measured population retains an interpretable primary texture. This can be tested with mineral boundaries, surrounding fabrics and comparison to less altered parts of the same stratigraphic section.

Every measurement needs a scale

A diameter reported in micrometres must be traceable to an image calibration. Suppose a scale bar represents 20 micrometres across 400 pixels. The scale equals 20 pixels per micrometre. An aggregate spanning 90 pixels has a calibrated diameter of 90/20 = 4.5 micrometres. The answer should say whether the value is a maximum Feret diameter, a manually estimated outline or another defined metric.

Do not copy pixel lengths into a spreadsheet and add a micrometre label after the fact. Different micrographs may have different magnifications, cropped scales or scanning resolutions. Software metadata may be unreliable after resampling. Preserve raw images, instrument settings and calibrations, and make the conversion reproducible for every image series. One mistaken scale can shift a whole population enough to change its apparent environmental category.

An irregular or elliptical outline has several possible diameters. Maximum Feret diameter measures the longest distance between parallel tangents; minimum Feret diameter measures another dimension. An area-equivalent circular diameter is defined by the diameter of a circle with the same measured area, or D = 2 times the square root of area divided by pi. These metrics are related but not identical. Declare one before comparing with a reference study.

For an illustrative outlined area of 19.635 square micrometres, D = 2 sqrt(19.635/pi), which is approximately 5.0 micrometres. This calculation measures a two-dimensional image footprint, not the true three-dimensional sphere diameter with guaranteed accuracy. Cross-sections through a sphere away from its centre look smaller than the sphere’s maximum diameter. Sectioning geometry adds a bias that a single precise area cannot remove.

If an aggregate’s long and short axes are 8 and 4 micrometres, an ellipse approximation has area pi times 4 times 2, or about 25.13 square micrometres. The area-equivalent circular diameter is sqrt(8 times 4), or approximately 5.66 micrometres. Using the maximum long axis would instead report 8.0. Both can be mathematically correct measurements of different shape properties; they should not be combined without conversion or a clear rationale.

Measurement repeatability matters when boundaries are fuzzy. A rough crystal edge, mineral coating or partially obscured aggregate can change the traced outline. Two analysts should independently measure a subset, with differences checked against the expected size range. A claimed change of 0.3 micrometres between groups may be smaller than the variation introduced by border tracing and pixel calibration.

Spatial resolution creates a lower detection limit. If the imaging method cannot reliably recognise aggregates below 2 micrometres, the observed distribution is left-truncated. Removing the smallest objects increases the measured mean and may narrow or distort the histogram. Include the minimum resolvable diameter and show how many borderline objects were excluded. A population that appears large may simply be missing its finest members.

A different bias occurs when the microscope field is chosen around obvious large or visually striking framboids. The analyst may unintentionally oversample large grains. Systematic or random field selection, documented exclusion rules and blinded counting where practical reduce this bias. The observation plan should be designed before the redox hypothesis is used to select attractive images.

From a list of sizes to a distribution

A distribution is more informative than a representative grain. At minimum, record number of measured aggregates, mean, median, spread, range, and ideally the underlying individual sizes or a histogram. The mean responds to unusually large values, while the median is less sensitive to a few outliers. A narrow distribution with small values has a different meaning from a broad population containing both tiny grains and large diagenetic aggregates.

Consider a constructed six-grain sample with diameters of 3, 4, 4, 5, 5 and 6 micrometres. Their sum is 27, so the arithmetic mean is 4.5 micrometres. The median is halfway between the third and fourth sorted values, also 4.5. The range is 3 to 6 micrometres. This is a compact set, but six observations are far too few to establish a robust geological population by themselves.

The sample standard deviation uses the sum of squared deviations divided by n minus one. For this constructed set, the squared deviations sum to 5.5 square micrometres. Divide by five and take the square root to obtain approximately 1.05 micrometres. A population standard deviation would instead use n in the denominator and give approximately 0.96. Specify which convention you used, especially when reproducing a graph from another study.

Now compare a second constructed six-grain sample: 3, 4, 5, 6, 10 and 20 micrometres. The mean is 8.0, the median is 5.5, and the sample standard deviation is about 6.36 micrometres. The lone 20-micrometre grain strongly raises mean and spread. A summary consisting only of mean 8 would hide the coexistence of several small grains with a potentially different large-grain component.

Histograms depend on bin width. One-micrometre bins can show detail but also noise; five-micrometre bins may combine distinct small subpopulations. Report the bin edges and compare results under reasonable bin choices. Kernel-density curves can help visualise structure, but smoothing does not invent additional observations. An elegant curve from ten grains is still supported by only ten grains.

The maximum diameter is sometimes used as an environmental clue, but maxima depend strongly on how many grains were examined. A sample of two hundred grains has a greater chance of including rare large grains than a sample of twenty. Comparing maxima without sample counts is unfair. Percentiles and the full distribution often provide a more stable basis for comparison, though they too require adequate sampling.

A sample’s small mean with broad variation may represent two populations rather than one intermediate condition. For example, tiny water-column particles could be mixed with large aggregates grown later in pores. A single Gaussian curve may be a poor model for such a mixture. Inspect histograms, image textures and stratigraphic relationships before interpreting the mean as a single homogeneous growth environment.

Another common confusion is between standard deviation and standard error. Standard deviation describes the spread among measured grains. Standard error of a mean, under basic independence assumptions, is approximately the sample standard deviation divided by the square root of the count. Increasing measured count can shrink standard error without reducing real grain-to-grain diversity. Both quantities have useful but different meanings.

Suppose a sample contains one hundred framboids with standard deviation 2 micrometres. A simple independent-grain estimate of the mean’s standard error is 2 divided by sqrt(100), or 0.2 micrometres. That figure does not include systematic boundary tracing, field-selection bias or correlation among grains from one microenvironment. A tiny reported error bar may describe mathematical precision rather than true environmental certainty.

Avoid pseudoreplication

A hundred grains in one microscope image are not necessarily one hundred independent geological observations. They may share a single localized chemical patch, burrow fill or nodule. To infer conditions across a bed, sample multiple independent fields and, where feasible, multiple powder or thin-section preparations. The appropriate unit of replication depends on the hypothesis: individual grain, image field, hand specimen, bed or core.

If all two hundred measurements come from one conspicuous cluster, averaging them may characterize that cluster precisely while telling little about the surrounding sediment. A hierarchical approach that groups grains by field and fields by sample can expose real spatial variation. Bootstrap resampling can be applied at the image or sample level when those are the independent units rather than treating every grain as exchangeable.

Sampling design matters for temporal claims as well. A bed with a high number of framboids may reflect a depositional event and not a sustained water-column condition. Multiple strata with consistent populations strengthen an interval-scale interpretation, especially if age control is independently established. The number of measured grains cannot substitute for the number of independently sampled geological horizons.

Two labs may produce different histograms from the same rock because they select fields differently, use different segmentation thresholds or classify overgrown grains differently. A reproducible study should share raw or annotated micrographs, detection thresholds and exclusion criteria. Interobserver agreement and repeat image analysis are useful checks. Simply reporting ‘framboids measured by SEM’ is not enough to understand the selection process.

What small and narrow distributions might mean

Modern euxinic analogues have yielded small, relatively uniform framboid populations consistent with nucleation and growth within a sulfide-bearing water column followed by settling. In the Black Sea comparison reported by Wilkin and colleagues, some Holocene populations had means near five micrometres and a large majority below seven micrometres. These are observations from particular environments, not a fixed worldwide size law.

A mechanistic explanation is that particles formed in the water column experience a limited growth time before sinking from the growth-favourable zone. Their size may remain comparatively small, and particles formed under similar conditions may have relatively narrow distributions. However, growth rates and access to iron and sulfide vary, and post-depositional processes can modify edges. A distribution supports the hypothesis only if those conditions are plausible.

In sediment pore waters, aggregates may experience different growth histories and access to local reaction zones. Some modern comparisons found larger or more variable framboids in deposits beneath oxygenated or low-oxygen water. This does not mean that every pore-water grain is large or that every small framboid formed in the water column. Formation times, saturation, nucleation surfaces and local methane-related reactions create overlap among populations.

A mean of five micrometres should not be converted to an ancient dissolved oxygen value. The proxy records a mineral growth process influenced by chemical conditions, not direct oxygen concentration. Small, tightly distributed grains may strengthen a euxinia interpretation where other evidence supports it. They do not independently establish that oxygen was exactly zero at every water depth or throughout the duration represented by the bed.

Large framboids can also develop near sulfate–methane transition zones and methane seep environments. Their growth may reflect sulfide generated during anaerobic oxidation of methane rather than routine organic matter sulfate reduction. That process introduces another mechanism affecting size and clustering. If a sediment contains methane-related carbonates, seep structures or distinctive isotope signals, a standard oxic-versus-euxinic size chart may need a different interpretive framework.

A narrow size distribution can be generated or preserved through physical sorting. Current transport may selectively move or retain particle sizes, while sediment reworking can mix old and young framboids. An apparently coherent population is not necessarily a direct water-column growth census. Sedimentary structures and the relation of pyrite to detrital grains help decide whether the original population has been transported.

A worked field of grains: the tempting false conclusion

Constructed microscopy sample A contains sixty intact aggregates measured across several documented fields, with a mean diameter of 4.8 micrometres and a sample standard deviation of 1.1. Almost all lie between 3 and 7 micrometres, and primary internal crystal textures are visible. This kind of distribution is compatible with some established euxinic analogues. The next question is not ‘which label do we print?’ but ‘what alternative formation and preservation histories must be checked?’

If the sample came from a finely laminated marine bed with high pyritization and independent evidence for sulfide-rich bottom water, the combination could support a water-column framboid interpretation. If instead the grains occur only inside a late fracture or a methane-related nodule, the same mean would not represent the primary water column. The microscopy and geological host determine how far the size statistic can be translated.

A sample mean is not enough to specify confidence. Report the measurement method, field count, grain count, standard deviation, median, detection limit and whether overgrown grains were excluded. Where a redox classification uses a published mean-versus-standard-deviation diagram, disclose which study and population it was calibrated from. Otherwise readers cannot tell whether the point’s position is meaningful or merely the product of selection.

Now imagine a second set of twenty grains from the same bed has a mean of 11 micrometres with broad spread and visible euhedral rims. Combining all eighty grains into one average could create an artificial middle category and erase evidence for a second generation of pyrite. A better analysis reports the populations separately, maps their textural contexts and tests whether the broad population represents late growth.

If a researcher excludes the twenty larger grains simply because they weaken an expected euxinia conclusion, that is selection bias. Exclusion is appropriate only when it follows independently stated textural criteria such as unmistakable secondary overgrowth or fracture-filling origin. The study must show the evidence for those criteria, not retroactively redesign the dataset to obtain a preferred environmental answer.

A second worked case: counting changes the mean

Constructed sample B has five measured diameters of 4, 5, 5, 6 and 20 micrometres. Their mean equals 8.0 micrometres. If the 20-micrometre grain was an authenticated later overgrowth, the primary four-grain group would have mean 5.0. Excluding it changes the result, but the exclusion must follow petrographic evidence rather than numerical inconvenience. The sample remains too small for strong environmental inference either way.

The example illustrates why individual micrographs and measurement records matter. An outlier can be physically meaningful: it may identify another episode of precipitation, sediment mixing or a distinct chemical setting. A statistical trimming rule that blindly discards the upper five percent may remove exactly the grains that reveal the sample’s complicated history. Quality control should distinguish measurement artifacts from geological heterogeneity.

A sensible workflow uses preliminary imaging to define classes, then a documented counting design to estimate each class’s size distribution. The analyst can report a primary-framboid population and a separate overgrown or late generation, each with its count and uncertainty. This approach respects real textures while avoiding the fiction of a single environmental measurement when several populations are present.

How to read a mean-versus-standard-deviation plot

Some paleoredox studies plot mean framboid diameter against standard deviation and compare the point with reference fields for different depositional conditions. The axes describe population statistics, not directly measured oxygen. A point’s location depends on the sample’s selection rules, number of observations and metric definition. Never reproduce a boundary from one study as a universal law without checking methodological compatibility.

A hypothetical point at mean 4.8 and standard deviation 1.1 micrometres appears in a smaller and narrower region than a point at mean 9 and standard deviation 5. Both points could be accurately calculated. Their meaning depends on whether the rock contains primary aggregates, late diagenetic material, methane-related pyrite, transport mixing or multiple chemical regimes. The crossplot begins a comparative inference; it does not complete the causal analysis.

If two samples overlap within uncertainty, do not announce a sharp redox boundary simply because one mean is slightly lower. Sampling noise and image-segmentation differences can exceed the gap. Repeat measurement, inspect class proportions and compare other proxies. A meaningful environmental shift should generally be supported by a coherent stratigraphic pattern rather than a single small movement on a classification figure.

An empty sample deserves special handling. No observed framboids may reflect oxygenated conditions, lack of iron or sulfur supply, poor preservation, weathering, insufficient sample area or imaging limitations. The correct result is ‘none detected under this procedure’, with detection effort and sensitivity stated. Absence becomes environmental evidence only when plausible alternative causes have been assessed.

Building a chain of independent evidence

Framboid microscopy works especially well alongside iron speciation because the two methods ask different though related questions. Reactive iron and pyritization ratios describe chemical inventories; microscopy describes particle texture and growth. Agreement can strengthen a hypothesis, but neither method is perfect. A high pyrite iron fraction does not certify that small framboids grew in the water column, and a small size population does not measure total iron supply.

Molybdenum and uranium enrichments can add constraints on reduction and sulfide-related reactions. Molybdenum may be retained in association with sulfide chemistry but is affected by seawater supply and particle shuttles. Uranium can accumulate in reducing pore waters. Thus trace metals can support or challenge a model without automatically pinpointing bottom-water conditions. Look for a physically coherent combination rather than counting independent ‘votes’ without examining shared processes.

Sulfur isotope measurements offer another view of microbial sulfate reduction and sulfur cycling. However, the isotope composition of pyrite depends not only on microbial fractionation but on diffusion, burial and sulfate availability. A small framboid and a particular sulfur isotope value should not be joined into a unique microbial pathway without testing their contexts. Multiple measurements can still be ambiguous if they reflect overlapping controls.

Sediment fabric matters because it links minerals to deposition. Laminated muds, trace fossils, event beds and erosional surfaces show how the sample formed and whether it was disturbed. A framboid size shift coincident with a major facies boundary may reflect a source change as well as oxygenation. Interpret the transition through sedimentology rather than treating all depth variation as redox.

Geographic replication is important for basin-scale reconstructions. A single core can document a local population but not prove a whole ocean’s conditions. Multiple water depths, basinal positions and comparable stratigraphic horizons help distinguish a local pore-water signature from an extensive water-column episode. The size of the claim should scale with the spatial and temporal coverage of the evidence.

Common errors and repairs

Error: measuring microcrystal diameters and comparing them with aggregate diameters from a published redox field. Repair: define whether D refers to the entire framboid, an individual crystallite or an overgrowth zone. Reanalyse images using a compatible metric. The same photograph can yield radically different numerical sizes depending on which boundary was traced.

Error: choosing the most impressive image as representative. Repair: select fields systematically or randomly and document the imaging coverage. Publish counts and distributions rather than relying on a single micrograph. An image is useful for showing texture; it is not a statistical description of the bed unless the sampling design supports representativeness.

Error: reporting a mean without spread or sample count. Repair: give median, standard deviation, range, percentiles where relevant, number of grains and number of independent fields. The distribution’s width can be as informative as its centre. A mean of five from three grains is not equivalent in confidence to a similar mean from a well-sampled series of fields.

Error: treating every small framboid as proof of euxinic bottom water. Repair: check pore-water formation, later growth, reworking, methane-related processes and comparison with modern reference populations. State the inference as conditional on primary preservation and compatible geological setting. Distinguish physical evidence from an inferred ancient water chemistry.

Error: deleting large outliers without reporting them. Repair: document all detections, classify genuine secondary textures with objective petrographic evidence, and show results with and without flagged grains. A multimodal distribution may be the most scientifically interesting outcome. Statistical convenience should not erase evidence for multiple generations or transport.

Error: ignoring calibration and detection limits. Repair: record the scale conversion, pixel resolution, boundary tracing rules and minimum resolvable aggregate size. A left-truncated histogram can shift its mean upward and change its spread. Recompute any compared datasets with consistent measurement definitions when possible.

Error: treating all grains within one image as independent replicates for a basin-scale claim. Repair: use a hierarchical sampling design with independent fields, beds and sites. Standard errors based on hundreds of correlated grains can be misleadingly small. Geographic and stratigraphic replicates support broader environmental inference more effectively than enormous counts from one cluster.

Independent transfer exercises with answers

Exercise one. A 30-micrometre scale bar covers 600 pixels. A framboid spans 110 pixels. Find its diameter. Answer: 600/30 = 20 pixels per micrometre, so 110/20 = 5.5 micrometres. The result is valid for that image and the chosen dimension, assuming the image has not been rescaled independently of its bar.

Exercise two. A circular aggregate covers 28.27 square micrometres in a calibrated image. Find its area-equivalent diameter. Answer: D = 2 sqrt(28.27/pi), approximately 6.0 micrometres. This is a two-dimensional area-based metric; it should not be mixed casually with a longest-axis metric in another dataset.

Exercise three. Diameters are 3, 4, 4, 5, 5 and 6 micrometres. Find mean and sample standard deviation. Answer: mean = 27/6 = 4.5. Sum of squared deviations = 5.5; divide by n minus one, or five, to get 1.1, then take the square root. Sample standard deviation is approximately 1.05 micrometres. Always identify which spread convention was used.

Exercise four. Two samples have equal means, but one has standard deviation 1 and the other 6 micrometres. Can they be treated as identical populations? Answer: no. Their spreads differ markedly and may reflect different growth histories, mixing, overgrowth or measurement problems. Examine histograms, individual images, sample counts and facies before inferring an environmental interpretation.

Exercise five. One sample contains twenty grains and another two hundred. Is their maximum diameter directly comparable? Answer: not without accounting for count and sampling effort. The larger sample has more opportunities to include a rare large grain. Compare fuller distributions and consider standardised sampling or probabilistic methods.

Exercise six. A dataset contains sixty small intact framboids and twenty large objects with clearly cross-cutting overgrowths. What should be reported? Answer: the full inventory plus petrographically justified subpopulations. Report exclusion criteria and results with and without the altered objects. Do not silently remove grains to force a preferred redox classification.

Exercise seven. A sample has standard deviation 2 micrometres for one hundred measured grains. Under independent assumptions, find the standard error of the mean. Answer: 2/sqrt(100) = 0.2 micrometres. This does not include calibration bias or spatial clustering and must not be mistaken for the full uncertainty of ancient environmental conditions.

Exercise eight. A study identifies tiny framboids only in a reworked turbidite layer. What is the leading interpretive hazard? Answer: the grains may have formed elsewhere and later been transported. Microscopic texture can be real while its location and timing are misplaced. Sedimentary provenance and stratigraphy are needed before connecting the population to local bottom-water chemistry.

Exercise nine. An image-measurement method misses every grain below two micrometres. What happens to the reported distribution? Answer: its lower tail is missing, potentially increasing the mean and altering the spread and percentiles. State the detection limit, and do not compare against a reference that counted smaller grains without noting the methodological difference.

Exercise ten. A paper declares a whole ocean permanently euxinic from small grains in two cores. What is missing? Answer: broader geographic coverage, adequate independent dating, preserved primary textures, sufficient sampling, multiple sedimentary environments and corroborating geochemical evidence. A local framboid result can support a regional hypothesis, but it does not establish global scale and duration by itself.

Frequently asked questions

Is there one universal micrometre cutoff for euxinia? No. Published reference environments show characteristic trends, and particular studies use interpretive size fields, but those fields depend on mineral formation, preservation, sampling and regional setting. Use appropriately matched reference populations and explain uncertainties rather than treating one cutoff as a physical law.

Can pyrite form under oxygenated seawater? Yes, in reducing sediment pore environments below an oxygenated overlying water column. Pyrite indicates that sulfide and iron reacted somewhere in the system; microscopy and context are needed to infer where that reaction happened. This is why pyrite presence alone is not proof of water-column euxinia.

Should I remove overgrown framboids? Report them, document the overgrowth evidence and analyse primary and altered populations separately when classification is defensible. The exclusion must be consistent and independently justified. The existence of a secondary generation is itself geological information that may matter for interpreting the sample.

Is a narrow distribution always better evidence than a broad one? It can be compatible with certain water-column growth settings, but physical sorting, method truncation or selective preservation can also narrow an observed population. A broad distribution may reveal mixed growth generations. Both require images, counts, appropriate statistics and environmental context.

What does a good final paragraph say? It reports the measured size population, the sampling design, the preserved textures and the calibrated comparisons. It then states which environment is consistent with the evidence and which alternative mechanisms remain possible. That is a useful scientific translation because readers can see the measurements, inference and limitations separately.