A postcard in a museum box shows a depot canopy and what might be a water tower. The back says “west side” in pencil. The database says “East Platform.” Both refer to the same building. The exhibit designer would like to show where the photographer stood so that today's visitors can see the difference.
What the designer actually has is a viewpoint problem. Identifying the depot might be straightforward: perhaps its sign is legible, and the same sign shows up in old maps. Identifying the camera position is harder. A structure can look similar from multiple corners. A rebuilt addition can make today's best match wrong for the period. An attractive modern street view can encourage experts to forget that the old picture may have been printed in reverse.
A robust way to approach an unattributed image of historic public infrastructure is to ask what fixed features should be visible together, in what order, from different proposed locations. Rank candidates against the visible relationships rather than trying to recognize a scene in one leap. The method is for archival photographs of public buildings and other open, historical landmarks. It is not a method for locating people, identifying private dwellings or publishing precise locations of sensitive sites. An exhibit often needs a block or public approach rather than an exact coordinate.
Separate the postcard's several clocks
A source can have more than one date. A photograph was taken at a time. A postcard might be printed later, sent later than that and acquired years after the sender wrote on it. A database record can give an estimated date or its own catalog revision. Keep those fields distinct. “Postmarked 1933” may support an outside limit on when that card could have been mailed or already existed, but it does not establish the time of day at which the image was exposed. A later paper caption can describe a location accurately, inaccurately or in terms no longer used by the local town.
Keep the physical front, back and their accession relationship together. Write down whose handwriting or note has been attributed, if anyone, rather than treating every inscription as contemporary. Locate any earlier print, negative or related exposure in the collection if one has been cataloged. A differently cropped print from the same original may show a corner or outbuilding removed from the familiar postcard. It's useful new visual evidence, but not an independent witness to the caption if produced from the same negative.
A replacement reproduction can introduce a new orientation, crop or contrast. Before asking what a left-to-right relationship means, look for a restrained orientation check already in the frame: unambiguous lettering, a known distinctive sign or a façade clearly compared with a separate period image. Not every apparent reversal is a print error; an object may be viewed from its other side. Do not automatically flip a picture until it fits your proposed place. Record both the orientation of the inspected source and which interpretation, if any, was verified.
Look for relationships that survive a different lens
Start with relatively fixed objects of known historical relevance: the building's main ridge, a water tank, a retaining wall, a public staircase, a roadway intersection or a bridge pier. Throw out portable features for the initial positional test. A wagon can turn around, passengers can move, and a sign on a fence can be relocated. Save the mobile details for the caption and the color of daily life. They rarely anchor the camera.
For each useful object, note its attachment and uncertainty. “Round tank behind depot roof” is easier to test than “water tower at northwest edge”; the latter already assigns your interpretation. Write “small cupola visually left of roof ridge” before deciding which direction left represents. Note whether a feature is seen in front of another or only left of it. Occlusion can be more useful than the amount of apparent overlap: if a roof is visibly in front of a tower base, the camera can't have seen the tower through that roof.
Use several clues that can fail independently. Roof form and roof window position may both come from the same assumption about the building; a tank across a roadway gives you a second anchor. A map and drawing might share source geometry; don't treat a repeated shape as independent corroboration until you know their origins. An independent record with a worse-looking scan can be more valuable than a near-perfect copy of the same source.
Perspective can make distance and size deceptive. Avoid measuring exact intervals in an ordinary street photograph without a documented method, camera geometry and suitable control. For a museum display, an ordered pattern of landmarks, their relative positions, and which obstruction hides which feature can do enough work. You can make a list or sketch rather than using a photogrammetry program. A student with paper can still ask a sound question: if the tower was behind the depot from this corner, which edge of the roof would it be near?
Choose maps and drawings that answer different questions
The USGS topoView help guide describes map year, edition, scale and downloadable formats. The historical topographic maps overview explains the historical collection. For an exhibit, a map from the relevant period might establish street layout, rails or the approximate relationship between landmarks. An absent detail on a small-scale topo is not evidence that the detail wasn't there. Read symbols, scale, explanatory notes and dates before deciding what a marked square or line means.
The USGS EROS aerial single-frame guide describes historical film frames and notes that the displayed original orientation may need rotation. It also discusses approximate geographic location derived from mosaics and flight-line maps. A frame may document a footprint or an open yard, but a thumbnail or rough location estimate is not a ready-made position for your ground camera. Confirm that the relevant structures actually appear in an image from a suitable period and that the image's orientation is understood before using it to support a direction. If you've found only a later aerial view, mark the comparison as later.
An architectural record can explain a façade in a different way. The Library of Congress's documentation note for the federal buildings, engineering and landscapes surveys describes measured drawings, photographs and written histories. If there is a period-adjacent survey for your public site, check the dates of documentation and fabric depicted, not just the original construction date. A measured drawing produced after restoration can be excellent at documenting that later configuration while being a misleading match for a photo from before the work.
An old bird's-eye view can surface a candidate arrangement. The Library of Congress Panoramic Maps collection guide describes them as perspective representations that are generally not to scale. Use these maps to see what the maker associated with a place and what to check next. Do not extract an exact camera position from a drawing made to show an appealing view of an entire town. A comparison is much stronger when three records are allowed to be different types of evidence rather than treated as photographs by another name.
Worked example: two approaches to Orchard Depot
This depot, its maps and the findings below are fictional. They demonstrate the method without identifying a real picture or exact viewpoint.
An exhibit team knows that the postcard depicts Orchard Depot. It sees five awning bays, a water tank to the left of the highest point on the depot roof, and a low stone wall that passes in front of the tank supports. The card was mailed in 1933. An archivist's later note calls it the east arrival platform. The museum has a hypothetical 1928 facilities plan and a 1936 aerial frame. Neither document identifies the ground photograph's viewpoint.
Draw the period footprint on paper. Mark the tank, wall, platform approaches and corners. On the 1928 plan the tank sits beyond the stone wall from the southwest approach. From the northeastern approach the tank is on the observer's side of the wall. Both approaches are consistent with seeing five awning bays; the depot has six on paper, but the postcard is cropped tightly after five. A count of five cannot disqualify either viewpoint unless you have evidence the image depicts the entire canopy.
| Observation from the example card | Southwest approach predicts | Northeast approach predicts |
|---|---|---|
| Tank base visually behind the stone wall | Consistent with plan | Tension: tank is on observer's side of wall in schematic |
| Tank left of the visible ridge peak | Consistent with team's upright sketch | Tension unless orientation or feature identity differs |
| Five visible awning bays, right end cropped | Possible | Possible |
| Postcard catalog says east platform | Contradiction if note identifies viewpoint correctly | Consistent with note, but its origin remains uncertain |
First use the obstruction relationship; the image shows the stone wall hiding part of the tank's support, not just near the tank in the rectangle. Then check the order in the same orientation in both candidates. Notice that the tank relationships depend on one identification. If a different structure is misread as the tank, both tests weaken. The research should find a third fixed comparison or clearly acknowledge that common dependency. The awning feature doesn't win the argument, but it remains useful context.
The hypothetical 1936 aerial confirms a structure in the approximate location of the mapped tank and an adjacent dark line that may be the wall. Because the source is later than the postcard's postmark and the feature recognition is uncertain, write that it supports the layout tentatively, not that it dates the photo. The archive finds a second period image of the depot, with the lettering preserved in normal reading order and a visible west-side stair at one end. The stair rail looks consistent with a cropped rail in the postcard. This is a new candidate test, not conclusive unless the stair layout is independently confirmed.
Now the museum can write a cautious caption: “The depot and its tank are visible. The relative positions are consistent with the southwest public approach shown on the 1928 plan, although the exact camera position and source of the later platform label are unverified.” The team can depict a broad wedge on its exhibit map instead of a pin. It doesn't need to conceal why the original catalog used a different word. Record both, and ask whether “east platform” could describe the named subject of the view rather than where the photographer stood.
The revision is small but meaningful. It doesn't ask visitors to trust a neat overlay. It shows what the team weighed and which assumption would need to change to overturn the result.
Let a rival viewpoint make a prediction
A good rival hypothesis deserves a way to win. If the northeast candidate were correct, what characteristic of the period scene would you expect in an uncropped version? Perhaps there would be a stair beyond the canopy, or the retaining wall would run behind the tower. State the predicted relationship first, then search the historical plans or fuller prints. If both candidates predict the same relationship, don't call it a test. Find a different feature or decline to rank them.
Also consider whether a wrong time period, a mirrored image or a drawing's inaccurate orientation would affect every prediction. You don't need to explain all known discrepancies before writing the caption. You do need to distinguish inconvenient evidence from an inapplicable comparison. A pencil label from an unknown hand is useful but may have low specific weight; a visible feature firmly documented in a period drawing may have higher weight; a popular modern map that has never dated the structure contributes almost nothing to a question about which expansion the camera saw.
Protect the audit from enthusiasm. Give the reviewer the two sketches with neutral labels rather than titling one “Correct west location.” Ask them to mark which features are definitely visible, plausible and not legible without explanation. Compare first impressions before showing the archive notes. If all the disagreements depend on one faint mark, find a cleaner source or narrow the claim. More reviewer agreement is not automatically more independent evidence, but separating observation from expectation reveals where the case actually turns.
Count the things you did not use
Keep a short exclusion note alongside the supporting clues. In the depot example, a small patch of dark roof may look like a shadow that points southwest, but without a dependable exposure time, season, surface orientation and record of how the image was processed, it belongs in the notes rather than the decision. A painted platform number is a useful hint if a contemporary plan has the same wording, but can be misleading if stations used the same naming scheme or numbers changed. Omitting a weak clue from the main figure doesn't mean suppressing it; log the candidate and reason for not relying on it.
This habit also helps later rechecks. Someone who recognizes the dark patch from a clearer scan should be able to add a better identification without recounting all the visual features from scratch. Save the rejected assumption, not just the rejected location.
Draw the uncertainty into the public figure
A suggested viewpoint can be a shaded sector rather than a sharp GPS point. A view direction can be approximate. A diagram reproduced from a dated plan can be paired with a note that the position is from that edition, not an assertion about the modern curb line. Keep the didactic picture and the sources in conversation: name the specific clues and how they appear in the drawing, rather than writing “the computer matched the photo” next to an exact-looking coordinate.
If it is safe and appropriate to look from a current public vantage point, a present-day repeat view can help an audience see a surviving landscape. Treat that as a comparison of two moments, not validation that the original photographer stood at an exact place. Public access can change independently of the historic scene. Don't tell exhibition visitors to step onto private land, operational infrastructure or road lanes to reproduce a view. Offer the diagram and the version of the photograph cleared for display.
Keep one plain record containing source image identifier and date limits, orientation evidence, each corroborating map or image and its vintage, candidate observations, strongest alternative, and caption decision. Put source citations next to the diagram. If new material arrives a year later, add a note rather than quietly altering the archive's record of what the older caption got from the evidence. This is particularly useful when the new material shows that part of an earlier conclusion was warranted: the building identification might survive even after the viewpoint changes.
An especially good label can teach a small piece of the method in one line: “The raised wall in front of the tank is the clue that points to this approach.” The next sentence can say what remains open. Visitors come away with a story about the depot, and a way to think about a different photograph they haven't seen before.
Record the picture you have, not the one you wish had survived
Before working with details, note whether the visible object is an original print, a copy, a postcard reproduction or an image whose source has not yet been found. The distinctions matter for viewpoint reasoning. A scanned postcard may omit details near the picture edge, and an exhibit reproduction may crop the only visible piece of the approach road. If you don't know how much has been lost, state that uncertainty rather than inferring that a missing sign must have been elsewhere in the landscape. A modern scanned margin and the physical border of the original object are not always the same boundary.
Use one stable identifier for the underlying archive item and a separate identifier for every derived image you interpret. A handwritten annotation visible on the back of a print can survive in a catalog page even when the version circulating on the web only shows the front. Conversely, a caption attached to a social-media crop may never have been part of the collection item. Preserve both when they are useful and label which is which. If the archive offers alternative exposures or different digital sizes, start with the version that gives the richest defensible evidence, and note the edition you used so another reviewer can repeat the comparison.
A reverse or reshot image is an especially important possibility when relationships depend on left and right. Legible text, reliably known building asymmetry, or other independent clues can help check orientation. No one of those signs is a magic reversal detector. A partly visible sign could be read incorrectly, and a documented building might be drawn in proposal form rather than as it existed that day. Do not diagnose a mirrored photograph because it disagrees with your favorite viewpoint. State explicitly whether the orientation was checked, how, and with what outcome before trusting a candidate that can only work if the picture is reversed.
For a fictional archive intake, assume the library has a postcard with a blank border, a later cropped scan in a clipping collection, and a seemingly fuller image from an unidentified donor. The postcard retains a postmark, the clipping prints a caption, and the donor image contains an extra piece of landscape at the edge. Treat all three as useful resources with different claims. Match prominent details to test whether they reproduce the same exposure. Then decide which evidence is documentary—date markings and catalog sources—and which is geometric—positions and overlap in the scene. A view can be the fuller of two pictures without becoming the best date witness or proving the caption.
Decide which relationships are robust enough to be a test
The simplest useful evidence is sometimes ordinal. One feature hides another; a water tower appears to the right of a corner; a path runs toward a gap between structures. Those relations may survive moderate uncertainty about focal length or print scale. An exact angular separation often does not. If the original photograph's camera and lens are unknown, do not convert a distance measured across the page directly to distance at the scene. The relative location of two surfaces tells a smaller and often more durable story.
Consider the relation “the signal post runs through the visible edge of the waiting-room roof.” That statement can mean several things: the post is in the foreground and obscures part of the roof, the two outlines meet at one visible point, or a dark mark on the print merely crosses both. Before presenting it as a foreground-order clue, inspect what trace actually ends and restarts, and distinguish a legible occlusion from apparent contact. Two objects framing the same patch of sky may be related visually without being close together on the ground. A sketch with named features and a question mark on the unreadable edge helps keep the uncertainty tied to the precise feature.
Horizon and elevation issues carry a different burden. A photograph at one unknown height can show more or less of a raised platform than a person standing at ordinary height nearby today. Sloping paths, embankments and the camera platform might explain some differences without changing the general approach direction. Use available evidence about stationary features and eye-level remarks as context, and avoid claiming an exact standing point or height when a broad alignment is all the record supports. The central task is an exhibit's viewpoint inference, not a surveyed camera reconstruction.
When comparing several clues, check for common dependencies. The visible edge of a wall and the visible tip of a gate may be manifestations of one hidden assumption about which object is in front. Do not treat them as fully independent confirmations. A second type of evidence, such as a separately established terrain constraint or the orientation of a documented façade, is more valuable than collecting five consequences of a single unknown. This is a judgment about evidence quality, not a quantitative confidence score. Name each relationship and which other relations would fail with it.
Put the maps and photo into the same historical question
A topographic sheet is capable of supplying broad alignments and terrain context; the method and intended scale of that sheet matter to finer claims. The USGS historical topographic map collection and topoView documentation give useful context on identifying historic sheets. Start with the edition likely relevant to the scene before comparing it to newer conditions. If a route appears on a revised map, state whether you determined it already existed in the target period or merely located it in a later depiction. Mapped presence is useful evidence of a source's information; it need not prove the route operated on the exposure day.
The inverse case is just as important. A small platform, shelter, spur or path missing from a published map might not be the kind of feature that edition was meant to show. Ask about the map's resolution and purpose, and avoid a false negative inference from customary omissions. This is where a project plan or engineering survey drawing may provide another view of the question. Read any drawing's title, date and description to see whether it describes an existing feature, a proposed design, or something reconstructed after the fact. The Library of Congress documentation note for HABS, HAER and HALS is useful for understanding several kinds of record and their roles, not as a blanket reason to treat every technical-looking drawing as measured on a known historical date.
A stylized map can still be welcome. Panoramic art might illustrate which building was locally prominent, the naming of a route, or the arrangement a maker chose to emphasize. It may intentionally enlarge one feature at the cost of formal scale. It is better to cite it for what it can contribute than to discard it because it is not a survey. Conversely, do not use the art's exact separation between two objects to eliminate a camera position. Keep the source's intended degree of geometry in the claim that depends on it.
If two maps disagree, catalogue the disagreement before picking a winner. One may show planned realignment, another a line in use, and a third current infrastructure that bears the same name. An exhibit can identify plausible historic approaches without needing to reconcile every stage in an entire network. Discuss whether the disputed map detail is actually necessary to distinguish the candidate views. If the inference survives without it, avoid making the map dispute an important premise. If it does not, move the disputed fact into the caption's uncertainty and let it motivate a future archive inquiry.
Test an additional viewpoint on paper without going near the railway
A synthetic farm-town photograph illustrates how to expand the test. The donated image shows a depot's broad side, a lamp silhouetted near its short end, and a lone tree rising behind roofline at the edge of the picture. A map reliably places a public spur road on one side and a historic pedestrian route elsewhere. Let those be alternatives despite the caption's informal direction. Draw a sheet with one simple top-down sketch for each answer. Do not add features absent from the evidence merely to make each sketch demonstrative. Instead mark roughly where the two documented structures ought to overlap in each alignment.
The road approach is initially appealing because its version of the depot wing matches the apparent side. But it requires the lamp to appear beyond the roof edge unless the lamp was temporarily not in the position shown by the later plan. The pedestrian candidate leaves more room for the observed overlap but requires a slightly raised look over an intervening verge. Neither adjustment is impossible; both reveal where the reasoning depends on unconfirmed context. Compare whether the lamp and the verge were reliably dated for the photo or whether either was imported from a later source. The most attractive of the two sketches does not eliminate the other by visual confidence alone.
Now introduce a fictional alternative donor copy in which the full sky is visible and the tree shows a second low branch crossing the very top of the roof. That line may align with both approaches if the tree's historical position is only approximately known. More tree detail has improved the image without providing an independent landmark. By contrast, a consistent coping wall that visibly hides the base of a fixed station stair can offer another useful relationship, if its documented physical position and date agree. The exercise shows why adding more annotated marks is not automatically adding more discriminating tests.
Have a reviewer state the strongest rival rather than trying to refute a deliberately weak straw candidate. Give that reviewer the same dated source packet and invite them to construct an actual plausible viewpoint that would preserve the important clues. If they can do it only by moving a documented wall, say that openly. If they can do it by changing a limb's guessed position, the evidence for a unique location is weaker than it initially seemed. The outcome may be a useful range of viewpoints and a narrower claim than the original donated caption. That is a successful result, not a half-completed test.
Keep this entirely on paper or in images when the physical approach is not safely public. A distinctive present-day photograph can be made from lawful public viewpoints if suitable, but the real archive problem does not require reoccupying the historic position. No reconstruction justifies walking on or near operating tracks or private land. If the relevant ground is no longer available to the public, put that fact in the exhibit and show the viewpoints with schematic projections. A well-marked sketch can teach the inference more plainly than a forced repeat photo.
Separate a finding, an illustration, and an invitation to investigate
A sturdy exhibit can divide its presentation into three short parts. The finding says what survives the comparison. The illustration points to which relationship makes it persuasive. The invitation explains what another record could change. For the farm-town example, the finding might say “Two documented approaches remain plausible; the available overlap favors a view from the former road side only if the lamp matches the historical plan.” The illustration can show precisely which clue is conditional. The invitation can ask for a dated plan or a securely attributable wider print, rather than requesting that visitors try to access a location.
A simpler reading level need not remove uncertainty. “This photo probably looks at the station from this part of town. We compared which nearby things hide others. A better dated picture might identify the exact spot.” The technical evidence is still there on an adjacent panel. Visitors come away with a concrete lesson about observation and more reasons to take the archive seriously. If space is very tight, the strongest qualification belongs immediately beside the candidate location rather than in a publication note no one will open.
Support clarity with alternate text and restoration notes. Describe the underlying object, orientation and the two relevant visible relationships without repeating unsupported assumptions about the location. If the displayed image was brightened or cropped for readability, record that brief change outside the main caption. Put the high-resolution evidence where an interested reader can see how the lines were interpreted, consistent with the collection's conditions for use. If low-contrast detail was restored for public display, keep the unchanged evidence available for comparison and note where the interpretation depends on the altered presentation.
An archive researcher can also leave room for community information that isn't yet evidence of where the camera stood. An oral account may preserve an older route name, a contemporaneous working practice, or a remembered original setting. Record the account with attribution, date and permission as appropriate; distinguish an eyewitness statement about an actual photograph from later recognition based on topic. A story can be valuable without directly proving the view. It can point to a candidate location or document that makes a later comparison stronger. Let it have that useful role rather than treating it either as a trump card or as an anecdote to discard automatically.
Keep future corrections tied to the claim they actually change
A late-arriving photograph may revise the apparent presence of an object without changing the identification of the depot. A newly found inventory may establish the station's local name but say nothing about the photographer. A contemporary drawing with clear provenance might rule out one candidate while leaving the date of the print unclear. Record what each source changed and what continued to hold. The goal is not to rewrite the whole evidence history whenever a new clue proves decisive in one part of the argument.
One reviewer check is particularly useful just before printing. Put aside the final label and hand someone the original view, the annotated copy, the source roles, the candidate list and the discarded clues. Ask them to write a caption from those parts. If their defensible result differs from the prepared copy, compare the claims side by side. You do not need to force their sentence to match yours; you need to know whether the difference arises from evidence, a hidden assumption, or the level of precision each person chose. This final check discourages the flattering version of the story from outrunning the work that got it onto the wall.
Keep a small comparison sheet rather than a score alone
If the candidates need a tidy summary, resist the convenience of one numerical score. A sketch that agrees with three unresolved tree marks but cannot explain a definitely visible building overlap may deserve less trust than the tidy total suggests. Use four simple statuses: compatible, inconsistent under the stated assumptions, unreadable from this source, and not relevant to the candidate. A short note explaining a specific incompatibility often carries more useful information than a percentage. It also prevents a blank cell from silently counting in favor of the candidate a project wants to promote.
One way to audit the sheet is to inspect what happens if its weakest premise fails. Remove the building plan that may be a proposal, or the caption that came only from a clipped reproduction. Ask whether the building identity, relative direction or exact corner remain justified independently. The pattern of surviving claims is an excellent outline for a public caption. A reviewer can clearly see whether a more precise location depends on one archival detail, and can spend their limited research time trying to confirm that detail rather than rebuilding every part of the working file.
In a guide designed for many different pictures, keep the column names broad enough for the next researcher: photo relationship, stationary evidence, age fit, candidate prediction and observed contradiction. Avoid a prescribed checklist that forces forests, farm towns and street stations to share the same landmarks. The method requires shared relationships, not shared objects. A modest sheet can keep changing while those distinctions remain useful.
AI disclosure: Drafted with AI assistance for Apiary. The Orchard Depot example, evidence and proposed captions are synthetic. The cited institutions document the actual map and image types; none verifies the fictional location.
Source notes
- U.S. Geological Survey: topoView help and metadata, Historical Topographic Maps, and EROS aerial single-frame records.
- Library of Congress: Documentation note for HABS/HAER/HALS and Panoramic Maps collection context.