Pinhole Portfolio Review and the Exposure Record as Data

The intention

Choose six to eight pictures from the season’s work, sequence them, and write a short statement about them. Then do the thing that makes this a science course: take every exposure log you have written since Part VI, put the numbers in one table, and extract two quantities that no manufacturer publishes — the effective speed of your paper in your processing, and the reciprocity trend of a material for which nobody prints an exponent.

The pictures are the visible output. The table is the one nobody else has.

Learning objectives

By the end of this session you should be able to:

• Select work against the intention written before the exposure rather than against how the picture turned out, and say why the two selections differ.
• Apply a fixed critique sequence to any photograph and produce a critique that names causes.
• Collate a season of logs and fit a personal effective speed and a reciprocity exponent from judged, ordinal densities, stating the resolution of the method and the uncertainty of both results.
• Separate a fault that belongs to the camera from one that belongs to the exposure, using the recurrence of the fault across the whole set as the evidence.
• Store and label negatives so that the log entry and the sheet cannot be separated, using enclosures that meet a named conservation standard.
• Write a plain portfolio statement with no marketing in it.

Prerequisites

Every assignment in this part, and its logs. If you did the assignments and did not keep the logs, do the collation with what you have and say in your statement that the data set is partial; that is a legitimate result, and it is a better lesson about record-keeping than any amount of being told. Panorama, Curved Plane and Multiple Pinholes immediately precedes this page. Part II’s measurement and uncertainty lesson owns the propagation this page uses.

Safety classification

No practical work and no chemistry. This is a reading, measuring and writing session at a bench with dry negatives on it, and it carries no safety level because there is nothing to classify.

What is not a hazard here, and why. Processed, washed and dried negatives are inert: the halide that was not developed has been dissolved out and washed away, and what remains is metallic silver in hardened gelatin on a resin-coated base. The one handling caution is for the negatives rather than for you — the Library of Congress’s handling guidance is freshly washed hands or clean lint-free cotton or inert plastic gloves, and never touching the image surface — and it is a conservation control, not a safety control. If you do choose to reshoot, the safety classification is the one on the assignment page you reshoot under, and it applies there in full.

Materials

Item	Quantity	Notes
Every negative made in this part	all of them	Including the rejects. Especially the rejects.
Every log sheet and prediction sheet	all of them	The prediction sheets are half the evidence.
A grey scale or step wedge print	1	The reference against which every density was judged.
Sleeving that passes the Photographic Activity Test	enough for the set	See Storing what you have made, below, for what that means and how to check.
A pencil, or a marking pen that passes the PAT	1	For labelling sleeves, never negatives.
A large flat surface or a wall	1	The sequencing board.
Squared paper or a spreadsheet	1	The collation table and the plot.

Estimated cost

Cost band £, and most of it is sleeving. Dated prices live in the laboratory planner.

The first pass: select against the intention

Lay every negative out, with its log entry beside it. Read the intention field before you look at the picture. Then sort into three piles.

Pile one: it did what the intention said. The prediction and the result agree, and the picture is the one you set out to make. These are the portfolio candidates.

Pile two: it is a good picture that was not the intention. The exposure went differently, somebody walked through the frame, the cloud arrived. Keep these separately, because they are evidence of something real, but do not let them into the first selection.

Pile three: it did not work. Also keep these. They carry most of the diagnostic information that What the data says about the camera will need.

!Why the selection is made against the written intention, and not against the picture

The obvious objection is that this is perverse: surely the best pictures should be selected, whatever you meant?

Two answers, and the second is the important one.

First, you can always add a pile-two picture at the end, and this page tells you to. The rule is not that lucky pictures are excluded; it is that they are identified as lucky.

Second, and this is the whole reason for the discipline: a portfolio selected only on how the pictures look tells you nothing you can act on. A portfolio selected against intention tells you your hit rate, which of the seven variables you can actually control, and which assignment you should repeat. A photographer who cannot repeat a good picture has a good picture; a photographer who can is a photographer. That is what the season of logs was for, and this is the page where it pays.

Then choose six to eight, at most one per assignment plus one alternate, and set them aside from everything else.

The critique sequence

A critique is worth something when it names causes. The sequence below is the course’s own, built from the arithmetic the part has taught rather than adapted from any published scheme, and it is fixed so that the same seven questions get asked of every picture and no comfortable one gets skipped.

The critique sequence: seven questions, in this order, for every picture

1. 1Intention — what the log said you were making; answered yes or no
2. 2Optics — name the fact: blur = d, cos⁴ θ at this corner angle, near-far at this focal distance, the cylindrical mapping
3. 3Exposure — how many stops out, judged against the grey scale, and inside or outside the band you predicted
4. 4Tone — predicted ranking against observed ranking; every inversion attributed
5. 5Time — what moved, for what fraction, and does the density agree with the presence arithmetic
6. 6Composition — the only question with no number; placed sixth so it cannot crowd out the six that have one
7. 7Execution — leaks, flare, fog, scratches, uneven development — faults of making, not decisions

The course's own sequence, derived from what this part teaches rather than adapted from a published critique scheme. Use it in order and in full; the value is in the questions you would otherwise not have asked.

Two rules about the writing itself. Every sentence that offers a judgement carries the evidence for it in the same sentence: not “the sky is too bright” but “the sky is at maximum density, which on a blue-sensitive material at this exposure is what the landscape page’s opening section predicted”. And the closing change is one change, expressed as a number: “start at 45 seconds rather than 65”, not “expose more carefully”.

Sequencing and pairing

Six to eight pictures in an order is not six to eight pictures. Three things to decide, in this order.

Openings and closings. The first picture teaches the viewer how to look at the rest; the last one is what they leave with. Neither should be the most complicated.

Rhythm. Alternate scale. A frame whose subject fills it, then one with a lot of air; a near-far-stretched foreground, then a flat compressed one. A sequence of seven pictures all made at f = 50 mm from half a metre reads as one picture seen seven times.

The flat-and-curved pair. This part gives you a specific opportunity nobody else’s portfolio has: you have two negatives of the same subject on two different projection surfaces, made minutes apart from one tripod position. Place them together, and let the sequence make the argument the panorama critique made in prose. If you have a paper-and-film pair from the landscape assignment, the same applies, and it argues about spectral response rather than about projection.

A pair is not the same as two pictures. Two similar pictures side by side compete; a genuine pair differs in exactly one controllable thing and is captioned, in the viewer’s head, by that difference. The test is whether you could state the difference in one clause — same subject, flat back and curved back; same subject, paper and film — without listing a second difference. If you have to say “and also the light had changed”, it is not a pair, it is two attempts, and it should be split up in the sequence rather than hung together.

A sequencing board, and what each position is doing

1. 1Open — legible at once; teaches the viewer how to read the rest
2. 2The scale contrast — filled frame, then open frame — do not ask the eye to do the same work twice
3. 3The pair — flat and curved of one subject, adjacent; the argument makes itself
4. 4The quiet one — placed where attention dips, which is about three-quarters of the way along
5. 5Close — what the viewer leaves with — chosen for what it says, not for what it cost

Lay it out physically and walk past it twice before fixing it. Alternates stay beside the board, not in the line.

The log as data

Now the part of the session that has nothing to do with taste.

Collate every paper exposure you have made since Part VI into one table, one row per sheet, five columns: the measured or estimated illuminance, the effective f-number, the time you computed from t = N²C/(E**S) with S = 3, the time you actually gave, and the judged result — thin, right, or dense — against the grey scale. If you did the assignments as written, you should have between sixty and eighty rows.

Two quantities come out of that table, and they come out of different features of it.

🧮DO THE MATHS: how a scatter of judged sheets separates into a speed and an exponent

Take the sheets you judged right. For each one, the time that worked, T_a, relates to the time you computed at ISO 3, T_c, by two effects at once: your paper’s speed is not exactly 3, and it suffers a reciprocity failure that no maker has quantified.

Write the speed error as a factor k = S_assumed/S_effective, and the reciprocity as the usual power law. Then

T_a = k · T_c^P , so log T_a = log k + P log T_c

Actual against computed time, with both effects present

That is a straight line on log-log axes, and the two effects sit in different places on it. The intercept is the speed error and the slope is the reciprocity exponent. Plot log₁₀ of the time that worked against log₁₀ of the time you computed, draw the best straight line you can through the points, and read both off.

Your effective speed is S_eff = 3 / k, where log k is the intercept. If the intercept is +0.30 — that is, everything took twice as long as computed — then k = 2 and your paper behaved as ISO 1.5 in your processing.

Your exponent is the slope. A slope of 1.0 would mean no reciprocity failure at all, which is not what anybody expects. A slope between about 1.1 and 1.6 puts your paper somewhere in the range the film makers publish for films.

The nomogram: lay your own points on it and read off the slope and the intercept

• P = 1.00 — no reciprocity failure
• P = 1.15
• P = 1.31 — ILFORD’s figure for HP5 PLUS, for scale
• P = 1.60 — a severe material
• P = 1.31 with a one-stop speed error (intercept +0.30)

Show the numbers behind this plot

Five straight lines on log-log axes, running from a computed exposure of one second at the left, where the logarithm is zero, to a computed exposure of a thousand seconds at the right, where it is three. Four of the lines pass through the origin and differ only in slope, each slope being a value of the reciprocity exponent. The lowest is a slope of one, the case of no reciprocity failure at all, rising from zero to three: a computed ten seconds would need ten seconds and a computed thousand would need a thousand. Above it a slope of one point one five reaches three point four five at the right-hand edge, so a computed thousand seconds would need two thousand eight hundred. Above that a slope of one point three one, the figure ILFORD publish for HP5 Plus, reaches three point nine three, so a computed thousand would need eight thousand five hundred. The steepest, a slope of one point six, reaches four point eight, so a computed thousand would need sixty-three thousand seconds, or seventeen and a half hours. The fifth line is different in kind: it is drawn parallel to the slope of one point three one but lifted by nought point three of a logarithmic unit, which is a factor of two, and it shows what a speed error looks like — every exposure twice as long as computed, with the same reciprocity behaviour. That is the whole method in one picture: a slope tells you about reciprocity and a vertical offset tells you about speed, and they are independent, so a scatter of points can give you both. A shaded horizontal band of plus or minus nought point one five of a logarithmic unit around the lowest line marks the resolution of judging density by eye against a grey scale, which is about half a stop, and it is drawn to show how far apart the points must be along the x-axis before a slope can be distinguished from the scatter.

Series	log₁₀ of the exposure you computed at ISO 3, seconds	log₁₀ of the exposure that actually worked, seconds
P = 1.00 — no reciprocity failure	0.00	0.00
P = 1.00 — no reciprocity failure	1.00	1.00
P = 1.00 — no reciprocity failure	2.00	2.00
P = 1.00 — no reciprocity failure	3.00	3.00
P = 1.15	0.00	0.00
P = 1.15	1.00	1.15
P = 1.15	2.00	2.30
P = 1.15	3.00	3.45
P = 1.31 — ILFORD’s figure for HP5 PLUS, for scale	0.00	0.00
P = 1.31 — ILFORD’s figure for HP5 PLUS, for scale	1.00	1.31
P = 1.31 — ILFORD’s figure for HP5 PLUS, for scale	2.00	2.62
P = 1.31 — ILFORD’s figure for HP5 PLUS, for scale	3.00	3.93
P = 1.60 — a severe material	0.00	0.00
P = 1.60 — a severe material	1.00	1.60
P = 1.60 — a severe material	2.00	3.20
P = 1.60 — a severe material	3.00	4.80
P = 1.31 with a one-stop speed error (intercept +0.30)	0.00	0.30
P = 1.31 with a one-stop speed error (intercept +0.30)	1.00	1.61
P = 1.31 with a one-stop speed error (intercept +0.30)	2.00	2.92
P = 1.31 with a one-stop speed error (intercept +0.30)	3.00	4.23

Lines computed from log Ta = log k + P log Tc; no data of the course's own is plotted, because the points are meant to be yours. Two things follow from the shaded band: your points must span at least two decades of computed time before a slope means anything, and no single sheet can tell you either quantity. The curve is drawn to show the shape, not measured from a real material. Your own materials will differ, and measuring them is what the sensitometry part of the course is for.

📓What this method can and cannot support

It rests on judged, ordinal densities. You looked at a sheet against a grey scale and called it thin, right or dense. Part VI’s own experiment page establishes that this resolves about half a stop at best, which is ±0.15 in log₁₀. Everything below follows from that one figure.

The speed you get is good to about half a stop. So if your intercept says ISO 1.5, the honest statement is “between about 1.1 and 2.1”, and if it says 4, the honest statement is “between about 2.8 and 5.7” — which, note, sits inside the range ILFORD themselves publish for these papers, an equivalent film ISO of 3 to 6. Your number is not more precise than theirs; what it is, is yours, for your developer, your dilution, your time and your light.

The exponent is much weaker, and needs a long lever arm. A slope fitted from two points one decade apart, each uncertain by ±0.15 in log₁₀, carries an uncertainty of 0.15√2/1 ≈ ±0.21. Spread the same two points over two decades and it halves to ±0.11. Fit a line through a dozen points evenly spread over two decades and it falls to about ±0.08. So the minimum useful data set is exposures spanning at least one hundredfold, which is exactly what this part gave you: still-life sheets of tens of minutes and landscape sheets of a few seconds. An exponent known to ±0.08 will tell you whether your paper is mild like a T-MAX film or severe like FOMAPAN; it will not tell you the second decimal place, and you should not write one down.

What it cannot do at all. It cannot give you a speed point in the ISO sense, which requires a measured density at a defined point on a measured curve, produced under a specified development. Parts XIII and XV do that with a densitometer and a step wedge, and the number they produce will not be this one. This is a working exposure index, honestly derived and honestly bounded, and calling it anything grander would be the kind of over-claim the course exists to avoid.

One further caution, from process control. ILFORD’s own argument for plotting a control result over time rather than judging a single one applies here: a single anomalous sheet is scatter, and only the trend across many is a result. Do not throw out a point because it is inconvenient; mark it, and see whether it has company.

Compare your two numbers with what Part VI told you to assume. Part VI said to take ISO 3 as a pessimistic starting point, expecting the bracket to move it, and it said flatly that no manufacturer publishes a reciprocity correction for photographic paper used as a camera negative and that the course would not invent one. You now have both numbers. Write them in the front of the log with the date and the developer, and use them from here on.

One sanity anchor. HARMAN publish guide times for their direct positive paper in a pinhole camera — one to two minutes in bright summer sunshine, up to about an hour for a lit interior. Their paper is slower than MULTIGRADE RC and their f-number is unstated, so nothing transfers arithmetically. But if your fitted speed makes a bright-sun exposure come out at ten seconds or at half an hour, look for the error before you believe the number.

What the data says about the camera

The logs also contain a second, quite different result, and it is found by looking for recurrence rather than by fitting anything.

Go through pile three — the failures — and every rejected sheet from piles one and two, and ask of each fault: does it appear on sheets from more than one session, with more than one plate, in more than one light? A fault that recurs across all of those is not an exposure error. It is the camera.

Fault that recurs across sessions	What it points to	The build change
One corner or edge consistently denser than the opposite one	A light leak at a specific joint, or a shifted plate	Re-run Part VI’s leak test with the suspect joint taped, and compare
Overall veiling greyness on sheets made towards bright light	Flare from the inside of the box or from the plate’s own bevel	Re-blacken the interior; check the plate’s front face is matt
Corners darker than cos⁴ predicts, at one focal distance only	The tunnel effect of a plate that is too thick for that focal distance	Thin the plate or countersink it, per Part VI
A soft band along one edge of every curved-back negative	The insert has sprung, so the paper is not at the focal distance there	Card tongues at both ends of the arc
A repeating mark in the same place on the sheet	Something physical in the camera: a burr, a fibre, a fleck of paint	Look for it. It is there
Density falling off towards one side on long exposures only	Sun on one wall of the box, warming it, or a slow leak that only shows in minutes	Shade the camera; repeat the leak test at the exposure length that shows it

This table is the reason the failures were kept. A single fogged corner is an anecdote; the same fogged corner on eleven sheets across four months is a measurement of your camera, and it is actionable in an afternoon with a roll of tape.

📓How recurrence tells you which part of the camera it is

The diagnostic power is not in noticing the fault. It is in sorting the sheets that show it against the configuration each was made in, which the log lets you do and memory does not.

Take a fogged corner appearing on eleven of about seventy sheets. Sort those eleven by focal distance. If all eleven were made at 80 or 120 mm and none at 50 mm, the leak is at a joint that only exists when an extension frame is fitted, and the frame’s shoulder is where to look. If the eleven are spread evenly across all three settings, the frame joints are innocent and the fault is in the body or the back, which is present in every configuration.

Now sort them by exposure length. A leak admits light at a rate, so its contribution grows with duration: a fault that appears only on sheets over about two minutes is a small leak, and one that appears on ten-second sheets as well is a large one. A fault whose severity does not scale with duration at all is not a leak — it is something in the optical path, or it happened outside the camera, in the holder or the changing bag.

And sort them by date. A fault that starts at a particular session and never stops is damage: the camera was dropped, a joint opened, felt came loose. A fault scattered through the whole season is a design or assembly defect that was always there.

Three sorts, ten minutes, and a fault with a location. This is what a season of logs buys that a season of pictures does not, and it is worth writing the sorts out even when you think you already know the answer, because the answer people already know is usually the last session they remember being annoyed by.

Reshoots

At most two, and each with one stated change, written down before you go out. A reshoot with three changes teaches nothing, because when it works you will not know which change did it.

Each reshoot is executed under its own assignment page, with that page’s safety classification, its materials and its log discipline in full. Reshooting is not a lighter version of the assignment; it is the assignment again with a hypothesis attached.

And two is a limit, not a target. If more than two of your eight pictures need remaking, the useful conclusion is not “remake them all” but “something systematic is wrong”, and What the data says about the camera is where to look for it.

Storing what you have made

The negatives leave this part and go to two places: to contact printing and enlarging in Parts XVIII and XIX, and to the capstone. They have to survive the journey and they have to arrive still attached to their logs.

Enclosures. The Library of Congress’s preservation guidance is unambiguous and repeated in three separate answers: all products used for storing photographs, and all materials in direct contact with a photograph, should pass the Photographic Activity Test (PAT), ISO 18916. The plastics they name as suitable for sleeves are polyester, polyethylene and polypropylene. The Image Permanence Institute’s negatives booklet cites the same test under its earlier number, ISO 14523:1999, alongside ISO 18902 for filing enclosures; the course has not verified the relationship between the two standard numbers beyond noting that both name the Photographic Activity Test, and the practical instruction is the same either way — buy sleeving whose supplier states that it passes the PAT, and if they cannot say so, do not buy it.

Marking. Label the sleeve, never the negative. The LOC note that permanent-ink pens which pass the PAT can be used on sleeves; a soft pencil on a paper enclosure is the other reliable route. Every sleeve carries the log entry’s identifier, and every log entry carries the sleeve’s, so that neither can be orphaned.

Conditions. Cool, dry, dark and — this is the part people skip — stable. The LOC give 18 °C as the highest recommended extended-term storage temperature for black-and-white prints and negatives on polyester base, and warn against daily fluctuations greater than ±2 °C. An interior cupboard beats a loft or a garage for exactly that reason: it is not colder, it is steadier. Part XII owns permanence properly; this is the minimum that keeps the work alive until then.

Labelling so that a negative and its log entry cannot be separated

1. 1Identifier — date and sheet number — the same string appears on the log row; this is the only field that must be unique
2. 2Assignment — which page it was made under, so a reshoot can find its parent
3. 3Configuration: f, plate letter, N — the compressed form the log already uses
4. 4Verdict at the time: thin / right / dense — so the collation table can be rebuilt from the sleeves alone
5. 5The negative itself — no writing at all — never mark a negative; the sleeve carries everything

Enclosures that pass the Photographic Activity Test (ISO 18916), in polyester, polyethylene or polypropylene, per the Library of Congress's preservation guidance. A pencil on paper or a PAT-passing pen on plastic is the marking.

The portfolio statement

Three hundred words at most, and the test is simple: every sentence either says what you did or says what you found. No sentence says how the pictures made anybody feel, and no sentence describes the work with an adjective that would fit any other body of work equally well.

A serviceable shape, which you should feel free to abandon once you can do better:

• What the pictures are of, in one sentence, plainly.
• What the instrument was — the camera you built, its focal distances and its apertures — in one sentence, because a pinhole photograph made with a bought camera and one made with a calculated one are different objects and it is honest to say which.
• What you were trying to find out, which is the intention field of seven logs, compressed.
• What you found, including the two numbers: your effective paper speed and your reciprocity exponent, with their uncertainties.
• What you would do next, in one sentence, as a specific test rather than an ambition.

✕The five sentences to strike out

Each of these appears in most first drafts. Each says nothing that could be false, which is what makes them worthless.

1. “I have always been fascinated by light.” Every photographer has. It distinguishes nothing.
2. “The pinhole sees differently from the eye.” So does every camera. Say how, with a number: 145.5° of horizontal field on a curved back, or a blur circle of 0.60 mm everywhere at once.
3. “I wanted to embrace the imperfections of the process.” If the softness was chosen, say which hole you fitted and what it cost you in exposure. If it was not chosen, say that instead.
4. “These images invite the viewer to slow down.” They do not invite anything. You made a four-minute exposure; say why you chose four minutes.
5. “The long exposure captures the passage of time.” It integrates the light over an interval. Say which interval, and what was in the frame for what fraction of it.

The test that catches all five: could this sentence appear, unchanged, in somebody else’s statement about entirely different pictures? If it could, it is not about your pictures.

What to submit

• Six to eight pictures, as negatives and working positives, in a fixed sequence.
• A critique for each, following the seven-step sequence, each closing with one numbered change.
• The collation table, every paper exposure from this part, with its five columns.
• The log-log plot, with your points on it, your fitted line, and both quantities read off with their uncertainties.
• Your effective paper speed and your reciprocity exponent, each with an uncertainty and a sentence saying what processing they apply to.
• The camera-fault table, with the recurrence evidence and the build changes you will make.
• The portfolio statement, under three hundred words, with none of the five sentences in it.
• The whole set sleeved and labelled, in PAT-passing enclosures, each sleeve tied to its log row.

What goes forward

To Parts XVIII and XIX, the negatives themselves, for contact printing and enlarging. Nothing in this part made a fine print, deliberately: a working positive exists to let you judge a negative, and judging a negative is not printing it.

To the capstone, the requirement to calculate a pinhole and expose negatives with it is met by this part’s work, and the portfolio pieces from it are candidates for the final selection. The capstone’s specification is the authority on which pieces count, and this page does not attempt to pre-empt it; what it hands over is a set of negatives, a set of logs, two measured numbers and a camera whose faults are now written down.

To the next season, one experiment. Look at your own collation table, find the largest source of uncertainty in it — it is usually the illuminance readings, occasionally the judged densities, and sometimes the times themselves on long exposures — and design the single test that would reduce it. Write it down. That is how a body of work compounds rather than merely accumulating.