Assignment: The Developer Laboratory Report

Somewhere in a box there are now about thirty strips of film, each one an answer to a question you asked on purpose. This assignment turns them into a document — and the test of the document is not whether it is tidy. It is whether a stranger who has never met you could read it, disagree with a conclusion, and know exactly which strip to re-measure.

That stranger is real. Part XXVII puts these same strips on a densitometer, and it will have your report and nothing else to tell it what they are.

What you are producing

One report, of perhaps two to four thousand words plus tables and plots, covering the three experiments of this part and the break/fix. Three hours is enough if the notebook was kept properly and is not enough if it was not, which is itself a finding worth recording in the first paragraph.

Part 1 — The structure, and why it is in this order

Six sections. The order is not a house style; each one is only interpretable in the light of the one before it, and readers who skip about are the readers who mistake a hypothesis for a conclusion.

1. Hypothesis. Copied from the notebook, verbatim, with the date it was written. Not rewritten. A hypothesis edited after the data is not a hypothesis, and the single most valuable line in many reports is “the prediction was wrong in the following way”.

2. Method. What was done, to the level of detail that lets it be repeated: the exposure batch block, every developer by formula version code, the volumes, the times, the temperatures with their ranges, the agitation script written out in words, and the reading method with its lighting. Deviations belong here and not in a footnote.

3. Data. Tables first, plots after, and nothing in this section that is an opinion. The measured values, the readings, the ranks, the deviations. If a strip was spoiled, it appears here with the reason.

4. Interpretation. The mechanism, and this is where Part VIII earns its place: what is happening at the crystal that would produce the pattern in section 3. Every claim carries its status — established chemistry, manufacturer’s statement, this course’s own measurement, or hypothesis — and the four are not interchangeable.

5. Conclusion. What you now believe, bounded by what the method can carry. Two or three sentences per experiment. If the honest answer is “the differences were inside my resolution limit”, that is a conclusion and it goes here.

6. Next changes. The proposal, versioned, with a prediction and a test. Part 5 of this page is entirely about that section.

📓The three sentences a report has to be able to write

Before you start, check that your notebook can supply all three. If it cannot, the gap is the first thing the report says.

• “The developer in cell A2 was D76b-EB-002, mixed on 14 August from 20.02 g of borax weighed on a balance checked against a 20 g reference the same morning, measured at pH 9.04 ± 0.2 at 20.1 °C.”
• “The strips were exposed in one sitting on 9 August, batch FP4-B1, lamp at 1.00 m, 8 s, wedge T2115, emulsion batch 20260614.”
• “The threshold step differed by two, and my resolution limit from the control set was one, so the difference is reported.”

Each of those is a sentence about a physical object with a history. None of them can be reconstructed afterwards from a memory of a good session.

Part 2 — Tabulating and plotting what a step wedge gives you

A step wedge gives positions, not densities. The arithmetic that follows from that sentence decides which plots are honest and which are not.

A Stouffer T2115 carries 21 steps at a 0.15 density increment, which is half a stop per step and three log units across the whole wedge. So a step number is a position on a logarithmic exposure scale, and the conversions that are legitimate are conversions along that axis.

Δ log H = 0.15 × Δ(steps), so the exposure ratio is 10^(0.15 × Δsteps)

A difference in threshold step, converted to exposure

Two steps is 0.30 log units, which is exactly one stop. That conversion is sound and you should use it, because “the threshold moved two steps” means nothing to a reader and “the bath needed one stop more exposure to reach threshold” means something to everybody.

What you may not do is turn a step count into a gradient. Average gradient is a ratio of a density difference to a log-exposure difference, and you have the denominator and not the numerator.

✕The seductive plot that must not be drawn

It is very tempting to plot step number against step number and call the slope a contrast. Two strips, two curves, one steeper: it looks like sensitometry and it is not.

Scale length in steps is a range, not a slope. A shorter scale means the tones ran out sooner, which usually means more contrast — but “usually” is carrying the whole claim, because the same scale length can come from a steep curve with a short straight line or a shallow one that hits the shoulder early. Without densities you cannot tell those apart.

So plot scale length as scale length, label the axis “steps from threshold to maximum”, and say in the caption that it is an inverse proxy for contrast and not a measurement of it. The proper version needs the densitometer, and the strips are archived for exactly that.

Three rules for every plot in the report.

Plot the measured value on the x-axis, never the intended one. The pH you measured, not the pH the arithmetic predicted. The concentration you actually made, not the one on the label — 0.49 g/L rather than 0.50, because the bromide went into 102 ml and not 100. This single habit is what makes a plot evidence rather than illustration.

Choose the axis scale from the design. The sulfite ladder is 10, 50 and 100 g/L, a tenfold range: on a linear axis two of the three points sit on top of each other and the shape is a lie. Use a logarithmic axis and say so. The bromide ladder at 0, 0.49, 0.98 and 1.96 g/L is linear-ish and includes a zero, so a linear axis is right and a logarithmic one is impossible.

Draw the resolution limit on the plot. Your control set gave you a spread in steps. Put it on every point as an error bar, or draw it once as a band and label it. A plot without it invites the reader to believe every wiggle.

One row of the data table, and everything that has to be in it

1. 1Identity — strip name, notch code, exposure batch — a reading with no strip attached is not a datum
2. 2The bath — formula version code, dilution as used, volume and the stock-equivalent it held
3. 3The variable — measured, with units — never the value you intended
4. 4The conditions — measured pH with its temperature and the electrode’s accuracy; time; temperature at start, middle and end; agitation script by name
5. 5The readings — random reading number, threshold step, scale length, fog rank, date and reader

Eleven columns is not bureaucracy. Cross out any one of them and name the question the table can no longer answer; that is the exercise, and every column survives it.

Part 3 — Comparing what you found with what the sources say

This is the section that separates a lab report from a diary, and it has a fixed shape: a table with four columns, one row per claim you can actually test.

The claim, and who makes it	What my strips showed	Agree?	What would settle it
“Too much alkali gives chemical fog” — Kodak, 1928	your fog ranking across the alkali ladder		a densitometer reading of base plus fog on the archived strips
“For greater sharpness, but with a slight increase in graininess, use a 1:1 dilution” — Kodak, D-76 sheet	your grain ranking at matched density		micrographs at matched density, or Part XXVII’s granularity readings
ID-11 at 1+3 recommended for maximum sharpness, PERCEPTOL stock for finest grain — ILFORD	your edge-target ranking		a densitometer trace across the edge
Bromide “retards rather than stops” — Hurter and Driffield, 1890	your B3 strip, and the spare developed for twice the time		a time series in the same bath
A pyrazolidone at one fifth to one sixth of metol’s weight gives similar characteristics — Kendall, 1941	your M against P, and MQ against PQ		matched-density comparison, which needs densities

Three rules for filling it in.

Name the status of every claim. Kodak’s 1928 sentence about alkali and fog is a manufacturer’s statement about behaviour; Hurter and Driffield’s is a published measurement; the regeneration account of superadditivity is, as Part VIII establishes, the standard explanation and not something any Tier-1 source in this course’s corpus states. Those are three different kinds of thing and a report that treats them alike has lost the distinction the whole course is built on.

A disagreement is a result. If your fog did not rise with pH, write that down as the finding it is, and then look for the reason in your own method before you look for it in Kodak’s. The most likely reason is that the effect was inside your resolution limit — which is a statement about your rig, and belongs in the report as one.

Never write “the literature says”. ILFORD’s own sheet models the right habit: it publishes its pH figures and then advises users to make their own control measurements from their own accurately mixed fresh solutions for later comparison. A manufacturer telling you to measure it yourself is the tone this section wants.

Part 4 — Uncertainty, and the sentences this method may not write

Every conclusion in the report is bounded by four kinds of uncertainty, and naming them is not modesty. It is what makes the rest of the report believable.

1. The resolution limit. The spread across your four identically developed control strips, in steps. Every comparison smaller than that is not reported as a result. Write the number at the top of the report.

2. Positions, not densities. You have threshold step, scale length and a fog ranking. You do not have a density, a gradient, a contrast index, a gamma, or a film speed in ISO — because every one of those is defined as a density or a ratio of densities.

3. One replicate, or none. Most cells in this part were run once. The solvent series contained a single duplicated condition, on purpose, and that pair is the only within-session estimate of repeatability you have. Two readings that differ by less than that pair’s own spread have not been separated.

4. The scanner’s numbers are relative. They are code values from an uncalibrated device with its own response curve, useful for ordering strips and useless as densities. Say which settings were used, or the numbers cannot even be compared with your own next session’s.

!Six sentences this report may not contain, and the honest version of each

• “The film’s speed in this developer is EI 64.” → “The threshold step moved two steps down against the control, which is one stop of exposure at the toe. A speed rating needs a density criterion and a densitometer.”
• “Contrast rose by 15 per cent.” → “The scale length shortened from 11 steps to 9. Scale length is an inverse proxy for contrast and not a measurement of it.”
• “The fog density was 0.12.” → “The fog on this strip ranked third of seven against clear fixed base, and was plainly visible where the control’s was not.”
• “Sulfite at 100 g/L gives finer grain than at 10 g/L.” → “At matched density, three readers ranked the 100 g/L strip finest and the 10 g/L strip coarsest, under a loupe at the magnification stated.”
• “The pH was 9.00.” → “The meter read 9.04 at 20.1 °C, with an electrode specified at ±0.2 units.”
• “Superadditivity was confirmed.” → “The hydroquinone-alone strip carried no density above base, so the additive prediction for the pair was the metol strip alone; the pair exceeded it by four steps of threshold. Additivity is refuted. The size of the excess needs densities.”

The pattern in all six is the same. The honest version names the reading, the instrument and the comparison; the dishonest one names a quantity nobody measured.

Part 5 — Formula versioning, and the proposal

This part of the course owns the versioning scheme, so this is where it is set out properly rather than used in passing. Part VIII’s assignment introduced the shape; what follows defines the fields, the rules and the reconstruction test.

Why a version code is not a name

“D-76” is the name of a formula. It is a published recipe: metol 2 g, sodium sulfite anhydrous 100 g, hydroquinone 5 g, borax 2 g, water to one litre. It names an intention.

“D76-EB-003” is the name of a litre. A particular litre, mixed by a particular person on a particular day from particular tubs, on a balance with its own errors, in water from a particular tap, and stored in a bottle with a known headspace since a known date.

Those are different kinds of object and a curve belongs to the second. Two litres of D-76 mixed a year apart are not interchangeable: one has borax weighed to 2.02 g and the other to 1.96; one was made with water that stood and one with water straight from the main; one is two months old in a half-empty bottle and Kodak’s own table says that is the end of its rated life. A result that cannot be tied to a specific mix is not evidence about a formula, and the entire diagnostic argument of the break/fix page turns on that sentence.

The three fields

STEM – INITIALS – SEQUENCE

The version code

The stem names the published formula the mix derives from, written without punctuation: D76, D23. It is a claim about parentage and it must be true — a mix that departs from the published quantities does not get the parent’s stem. It gets a variant stem: the parent plus a variant marker, D76b, D76c, declared once in a variant record that names the parent and the single change. This is the same distinction the course’s formulary draws with its status: variant and basedOn fields, and the notebook uses it for the same reason: so that a reader meeting the code for the first time knows immediately what it is not.

The initials name the person who mixed it. This is not vanity and not credit. Two people weigh differently, read a meniscus differently and stir for different lengths of time, and a systematic error belongs to a hand. When a series of results drifts, “who mixed it” is one of the first columns worth sorting on.

The sequence is three digits, counted per person per stem, in the order the batches were mixed: 001, 002, 003. Never reused and never renumbered. A renumbered batch orphans every strip, plot and conclusion that cited the old number, and a reused number silently merges two different litres into one apparent object. If a batch is spoiled before use, its number is spent; the next one is the next number.

What the code points to

A version code is an index key, and it is worth nothing without the record it indexes. The record is written once, when the batch is made, and it carries:

Field	Why a later reader needs it
The formula and the source, by document and section	So the intention can be checked against the recipe
Every mass actually weighed, to the resolution of the balance	The real composition, not the intended one
The balance used, and when it was last checked against a reference	Systematic error lives here and travels with the hand
The water: source, whether it stood before use, temperature at each dissolving stage	Aeration and hardness both act; Kodak’s dissolving temperatures are part of the formula
The dissolving order, and any departure with its reason	The order is chemistry: metol before sulfite, because metol is only slightly soluble in sulfite without alkali
Final volume, and how it was determined	A 3 per cent volume error is a 3 per cent concentration error
Measured pH with the temperature of the measurement, and the instrument’s stated accuracy	A pH without its temperature and its uncertainty is not a measurement
Colour and clarity when fresh	The zero reading for every future look at the bottle
Date, time, bottle and headspace	Headspace sets shelf life; Kodak gives six months full against two half-filled
Who mixed it	Because the initials in the code have to mean something

A new version is required whenever anything in that list changes. A different tub of borax is a new version. Tap water instead of stood water is a new version. Mixing at 35 °C instead of 45 is a new version. This is not fussiness: each of those is a change the strips cannot tell you about afterwards.

📓Dilutions, and the one place the code needs a suffix

A working solution is not a new formula and does not get a new stem. It gets the stock’s code plus the dilution and the date it was made:

D76-EB-003 @ 1+3, 2026-09-04

Two reasons the date is not optional. Kodak’s instruction for the 1:1 dilution is to dilute it just before use and discard it after one batch, and ILFORD’s is that diluted developers are not kept beyond 24 hours — so a diluted solution has a shelf life measured in hours and its age is part of its identity. And the stock it came from is ageing on its own clock underneath, which the stock code carries and the working solution does not.

The archive slip on every strip in this part therefore carries the stock code and the dilution and the volume in the vessel, because a strip developed in 80 ml of 1+3 and one developed in 160 ml of the same 1+3 did not have the same amount of developing agent available to them.

The reconstruction test

Here is how to tell whether a record is adequate, and it takes ten seconds.

Hand it to a stranger who has your balance and your tap. Can they make the same litre?

If any answer is “they would have to ask me”, the record has a hole in it and the hole has a name. They would have to ask which sulfite — anhydrous or heptahydrate. They would have to ask how warm the water was. They would have to ask whether the 2 g of borax was weighed or spooned. Every one of those is a field in the table above, and the table is simply the list of questions a stranger asks.

Apply the test to your own three sessions before you write the proposal. The gaps you find are the honest first paragraph of the report’s Method section.

The proposal: one variable, a prediction, and a test

The last section of the report proposes one modified developer, and it is graded on its discipline rather than on its ambition.

A proposal has five parts and no more.

1. The version code and the parent, e.g. “D76b-EB-001, based on Kodak D-76 as published in the 1949 Kodak Limited handbook.”
2. The single change, as a quantity, e.g. “borax raised from 2.0 to 20.0 g/L, which is the quantity Kodak specifies in the D-76R replenisher.” One change. A proposal with two changes is a proposal that cannot be evaluated.
3. The evidence it rests on, from your own strips, by cell name: “the activity series showed the metol-sulfite base gaining n steps of threshold and rising m ranks of fog between the unmodified base and the base with 20 g/L of borax.”
4. The prediction, stated so it can fail. “Against D-76 at the same time and temperature, this variant should shorten the scale length by at least two steps and raise the fog rank by at least one. If the scale length moves by less than my resolution limit, the prediction is wrong.”
5. The test, in one paragraph: how many strips, in what conditions, against which control, and what reading decides it.

!A proposal is not a formula, and the difference is the whole of Rule 1

Nothing you propose here is a developer this course recommends, and nothing in the report should be written as though it were. It is a hypothesis with a recipe attached. It becomes a formula when it has been mixed, run against a control, read blind and measured — and even then it is a formula for your film, your water and your rig.

The strongest proposals in this part are usually the least dramatic ones, because they take a quantity a manufacturer already publishes somewhere and move it into a formula where it is not. Twenty grams of borax per litre is not invented: it is what Kodak puts in the D-76R replenisher. Bromide at 1.9 g/L is not invented: it is DK-15’s level. A proposal built from published quantities can be defended even when the result is disappointing, and one built from a number you liked the look of cannot be defended at all.

Part 6 — Art track: the negative you now want

One paragraph, and it is not decoration. It is the section that turns thirty strips into a decision.

Write it in this order. What you photograph, in a phrase. What has been going wrong with the negatives you actually print — highlights blocking, shadows sitting on a veil, grain too visible at the enlargement you use, edges mushy. Which strip in your archive looks like the answer, by cell name. What you will change, in the language of this part: an alkali, a restrainer, a solvent level, a dilution, an agitation script, or an agent. And what it will cost, because every lever in Part IX costs something and a paragraph that claims a free improvement has not understood the part.

Kodak’s 1928 primer is still the best short vocabulary for the last two sentences: an Elon image comes up quickly and gains density slowly, a hydroquinone image comes up slowly and gains density steadily and rapidly. Between those two behaviours and the three levers of the solvent series there is enough language to say what you want without inventing any.

What to submit

• The report, in the six sections of Part 1, with the resolution limit stated at the top.
• The data tables, one row per strip, with the eleven columns of the drawing in Part 2.
• At least four plots: threshold against measured pH; threshold against bromide concentration; a sulfite plot on a logarithmic axis; and one plot of your own choosing that you think the data supports. Every one drawn against measured values, with the resolution limit shown.
• The comparison table of Part 3, filled in, with the status of each claim named.
• The uncertainty section, naming your four bounds and listing at least two conclusions you would have liked to draw and could not.
• The versioned proposal, in the five parts of Part 5.
• The art-track paragraph.
• The archive index: every strip, its sleeve, its code, and the batch it came from — the document Part XXVII will actually open first.

📓The one-page front sheet, which is the part that gets read

Whatever else the report contains, put a single page at the front carrying: the exposure batch name and date; the resolution limit in steps; the list of formula version codes used, with their mixing dates; the number of strips archived and where they are; and the three findings you would defend.

Everything else in the report exists to support that page. If the front sheet cannot be written, the report is not finished — and in a year, when you have forgotten the whole of it, the front sheet is what you will read.

Recap

• A report is a document a stranger can act on, and the stranger is Part XXVII with a densitometer.
• The six sections are in causal order, and the hypothesis is copied from the notebook rather than rewritten to match the result.
• A step wedge gives positions. Convert them to stops, never to gradients, and label a scale length as an inverse proxy for contrast rather than a measurement of it.
• Plot measured values, choose the axis from the design, and draw the resolution limit on every plot.
• Compare with sources in a four-column table, naming the status of each claim; a disagreement is a result, and the first place to look for its cause is your own method.
• Four bounds hold every conclusion: the resolution limit, positions rather than densities, one replicate, and an uncalibrated scanner.
• A version code names a litre, not a recipe — stem, initials, sequence — and it is worth nothing without the mixing record it indexes. The test is whether a stranger with your balance could make the same litre.
• A proposal is one change, one prediction that can fail, and one test, built where possible from a quantity somebody has already published.

Check your understanding

Question 1. Your threshold step moved from 17 to 15 between two cells. What may you write, and what may you not?

• That the film speed fell by one stop
• That the threshold moved two steps, which on a 0.15 increment wedge is 0.30 log units, or one stop of exposure at the toe — but not that the film speed changed, because a speed rating needs a density criterion and a densitometer
• That contrast fell by two steps
• Nothing, because step numbers cannot be converted

Show the answer and why

Answer: That the threshold moved two steps, which on a 0.15 increment wedge is 0.30 log units, or one stop of exposure at the toe — but not that the film speed changed, because a speed rating needs a density criterion and a densitometer

The wedge’s increment is a published property of the part: 0.15 density units per step, half a stop each, three log units across 21 steps. So a difference in step number converts exactly along the exposure axis, and that conversion is worth making because "one stop more exposure to reach threshold" is a sentence a reader can act on. What it is not is a speed rating: ISO speed is defined by reaching a stated density above base, and a density is precisely the thing a step reading does not supply.

Question 2. Why is the sulfite series plotted on a logarithmic x-axis and the bromide series on a linear one?

• Convention: concentrations are always plotted logarithmically
• The sulfite levels span a tenfold range, so a linear axis crowds two of the three points together and misrepresents the shape; the bromide series is a near-linear ladder that includes a genuine zero, which a logarithmic axis cannot show at all
• Because sulfite acts logarithmically and bromide does not
• To match the wedge’s logarithmic steps

Show the answer and why

Answer: The sulfite levels span a tenfold range, so a linear axis crowds two of the three points together and misrepresents the shape; the bromide series is a near-linear ladder that includes a genuine zero, which a logarithmic axis cannot show at all

The axis is chosen from the design of the series, not from the chemistry. Ten, fifty and a hundred grams per litre is a ratio ladder and reads properly only on a ratio axis. Zero, 0.49, 0.98 and 1.96 grams per litre includes a zero, and zero has no place on a logarithmic axis, so the choice is made for you. Getting this wrong is one of the commonest ways an honest data set is made to look like a different shape than it is.

Question 3. What is the difference between "D-76" and "D76-EB-003"?

• None; the second is just a longer way of writing the first
• The first names a published recipe, an intention; the second names one physical litre with a history — a hand, a balance, a water, a date and a headspace — and a curve belongs to the litre
• The second is a variant formula
• The second indicates the dilution used

Show the answer and why

Answer: The first names a published recipe, an intention; the second names one physical litre with a history — a hand, a balance, a water, a date and a headspace — and a curve belongs to the litre

The distinction is what makes a result attributable. Two litres of "D-76" mixed a year apart differ in the masses actually weighed, the balance that weighed them, the water, the dissolving temperatures and the bottle’s headspace, and Kodak’s own storage table says a half-filled bottle has a third of a full one’s rated life. Any of those could produce the difference between two strips. Tying every strip to a version code, and the code to a mixing record, is what turns "the developer" into an object with a testable history.

Question 4. A student proposes "D76b-EB-001: borax raised to 20 g/L and sulfite reduced to 50 g/L, for a sharper, more contrasty negative." What is wrong with it as a proposal?

• The quantities are not published anywhere
• It changes two variables, so no outcome can be attributed: if the strip is sharper and more contrasty, the report cannot say which change did which, and if it is neither, both changes are implicated
• The version code is malformed
• The prediction is too specific

Show the answer and why

Answer: It changes two variables, so no outcome can be attributed: if the strip is sharper and more contrasty, the report cannot say which change did which, and if it is neither, both changes are implicated

It fails the single rule the whole part exists to teach, and it fails it in the section that is meant to demonstrate the rule has been learned. Both quantities are in fact defensible on their own — 20 g/L of borax is the D-76R replenisher level and 50 g/L of sulfite is what D-76 at 1+1 already contains — so the proposal is not reckless about provenance. It is reckless about attribution, and the fix is to propose one of them and name the other as the next experiment.

Question 5. Your fog ranking shows no rise across the alkali ladder, contradicting Kodak’s statement that too much alkali gives chemical fog. What goes in the report?

• Nothing; a result that contradicts a manufacturer is probably an error
• The disagreement, as a finding, together with the first candidate explanation — that the effect was inside your resolution limit — and the test that would settle it, which is a densitometer reading of base plus fog on the archived strips
• A note that Kodak’s statement applies only to paper developers
• A recommendation to repeat the experiment until it agrees

Show the answer and why

Answer: The disagreement, as a finding, together with the first candidate explanation — that the effect was inside your resolution limit — and the test that would settle it, which is a densitometer reading of base plus fog on the archived strips

A disagreement is data, and suppressing it is the one thing a lab report must never do. But the reporting has an order: look for the cause in your own method before you look for it in the source, because the commonest explanation for "no effect" in this part is an effect smaller than a by-eye reading can resolve. That is a statement about your rig and belongs in the report as one — and because the strips are archived, the disagreement is not left hanging; it is handed forward with the test that decides it.

Question 6. Why does the archive slip carry the volume in the vessel as well as the dilution?

• To help estimate waste
• Because two strips developed in the same 1+3 solution but in 80 ml and 160 ml did not have the same amount of developing agent available, and Kodak’s minimum-volume figure is about agent per unit of film area rather than about covering the film
• Because the vessel affects agitation
• Because volume is needed to compute the pH

Show the answer and why

Answer: Because two strips developed in the same 1+3 solution but in 80 ml and 160 ml did not have the same amount of developing agent available, and Kodak’s minimum-volume figure is about agent per unit of film area rather than about covering the film

Kodak publishes one 135-36 roll — 80 square inches — in 473 ml of 1:1 developer, which works out at about 2.96 ml of stock per square inch. A 135 mm strip of 35 mm film is 7.32 square inches and therefore needs about 21.7 ml of stock however much water surrounds it. At 1+3, 80 ml of working solution contains only 20 ml of stock, below that figure, and the resulting shortfall would look exactly like an effect of dilution — which is the variable under test. The volume is part of the strip’s identity for the same reason the dilution is.

Check my answersClear