"He that doth rush to the forge without first considering his metal
shall produce nothing but noise and smoke, and count himself diligent for the effort."
Ars Magicae Fundamentalis attributed to Gerhard of Cologne
The first of November arrived with no ceremony. The dormitory was still dark when I woke, the same deep underground dark that made it impossible to tell five in the morning from noon, the air close and smelling of the tallow from the previous night's candles. Thomas was asleep. I lay still for a moment, listening to the breathing of the room and the particular silence of the castle beneath it, and then I got up, dressed in the dark, and went to pray.
I did not think about what I had seen the night before or rather, I thought about it once and set it aside. There was nothing to be done about it this morning, and nothing to be done about it the next morning either. I had prayed over it thoroughly the night before, thorough enough that I had run out of words somewhere around Job, and whatever I still carried from it would carry itself until time saw to it. That was how it worked.
Breakfast was oats with no salt, same as every other morning.
The week resumed. Classes, meals, the work assignment on Saturday, the common room in the evenings with its fire and familiar voices. I took part in all of it, ate at the same table, walked the same corridors. Thomas said something funny at supper on the fourth of November and I laughed, because it was genuinely funny. Eleanor asked about the Transfiguration reading and I answered her. From the outside, nothing had changed.
What had changed was where my free hours went.
Before All Hallows' Even I had spent them in the common room reading sometimes, more often just present while the others talked, which had been pleasant enough in its way. After, I spent them in the empty classroom on the third floor that no one seemed to use after midday, with a small knife borrowed from the kitchen, several handfuls of scrap pine from the carpentry work assignment, and my runes text open on the desk beside me. The room had a single window that faced north and admitted gray light through the afternoon, enough to work by without a candle until the last hour, when I would light the stub I kept in my coat pocket and work until the classroom grew too cold or the light had burned down past usefulness.
Nobody asked. Margaret gave me a look the second afternoon that told me she understood, and left it there. Thomas and Eleanor noticed I was gone more, asked after me once, and when I told them I was working on something they accepted it and forgot about it within the week, the way elevenyearolds forget most things that don't directly concern them. There was no falling out. There was just the common room in the evenings and the empty classroom in the afternoons, and the work I had set myself there.
I had read the runic principles three times by that point, carefully, with the book flat on the desk and my notes open beside it, and the more I read them the more they resembled something I already understood from a different direction.
The Second Principle, Ordered Flow, stated that magic travels through runes as water through channels, that the general form was Power to Condition to Operation to Result, and that without ordered flow the magic disperses or collapses. The author used water as his analogy throughout rivers, vessels, liquid seeking its own level. The Third Principle, Containment, required that all runic works be enclosed within some geometric boundary: a circle for stable spellwork, a square for binding, a triangle for amplification. Open systems, the text warned, allowed magical influence to diffuse outward without producing concentrated effect.
Water worked well enough as an analogy. But reading it, I thought of something else: current through a wire.
The rune was a conductor. The carved channel between runes was a conductor. The power rune at the start of the sequence was the voltage source, supplying aether the way a battery supplies potential difference, and the aether flowing through the inscribed lines from one symbol to the next was current moving from high potential to low along the path of least resistance. The action rune at the end of the sequence was the load the component that converted the supplied energy into observable effect. Containment was a closed circuit. Without closure the current had nowhere to return; without a return path, no stable flow established itself and no work was done. Even the Third Principle's table of enclosure shapes mapped roughly to circuit topology: a circle was a loop, a square was a bounded box, a triangle was a directed network with a defined apex.
I held that comparison for a while before I picked up a knife. I was fairly certain there were limits to how far it held, since magic was not electricity and pine was not copper, but the structural logic seemed consistent enough to give me something to think against. The analogy would break somewhere analogies always did but it gave me a place to start, and a place to start was worth more at this stage than any amount of speculation about where it might eventually fail.
The first attempt was simple by design. Three runes carved in sequence into a piece of pine roughly the size of my palm, the grain running lengthwise, no enclosure. Fehu at the start the text classified it as a raw power rune, the source of magical energy with no specific directional quality. Kenaz in the middle, associated with fire and heat generation. Thurisaz at the terminal end for force, because I did not want to set anything on fire while sitting alone in a room with no one who knew I was there. I wanted an output I could observe without evacuating the building.
The channels connecting the runes were straight cuts, made with the knife tip dragged at a consistent depth along a line I had scored with a thumbnail first. The cuts were perhaps three inches long each, the width of a shallow scratch. I had spent most of the previous afternoon deciding whether to vary any of this in a first attempt and concluded that varying nothing was the correct choice. Establish a baseline first. Change one thing at a time after.
"Fehu sources, Kenaz conditions, Thurisaz outputs," I said to the empty room, which was the habit I was developing of talking through the setup before activating anything, the same way I used to narrate a circuit analysis before committing a result to paper. It helped me catch errors. "Simple series. No branching. No containment that comes later."
I pressed my thumb against the Fehu rune and waited.
The wood beneath my thumb was cool and slightly rough where I had not sanded the surface, the carved lines sharper than the surrounding grain. For a full five or six seconds nothing happened, and I had time to consider whether I was doing something wrong before the sensation arrived. The Kenaz rune went warm. Not a spreading warmth, not a general rise in temperature across the board it was local, concentrated in the carved symbol itself, diffusing slightly outward into the surrounding grain in a radius perhaps the width of two fingers, but sharper and more distinct at the carving than anywhere around it. I counted the duration against my breathing. Three or four seconds. Then the warmth receded, not all at once but gradually, and by the time I had counted to eight it was gone.
The Thurisaz rune at the terminal end showed nothing. No warmth, no vibration, no change I could detect by touch or sight. I pressed my thumb against it directly for a moment, in case the effect had been too faint to perceive from the Fehu position, and felt only wood.
I set the board down and looked at it. No burn mark at the Kenaz position, which I had half expected after the warmth. No darkening of the grain, no change in color along the channels. The channels themselves were exactly as I had cut them. Whatever had moved through the board if anything had had left no physical trace beyond the warmth that was already gone.
I wrote in my notebook: First activation. Thermal response at middle rune, Kenaz. Duration three to four seconds by breath count. No visible discoloration or marking. No detectable response at terminal rune, Thurisaz. Sequence appears incomplete either flow terminated at Kenaz, or Thurisaz received insufficient energy to produce observable effect. Then underneath: Possible circuit analogy break between Kenaz and Thurisaz, or load resistance at terminal rune exceeds available current. Or: analogy fails here and middlerune termination is correct behavior I don't yet understand.
That last note was the honest one. Every engineer's instinct said label the failure and move on, but this was not a system I had designed from known principles it was a system I was inferring from observed results, and the difference mattered. The same result that looked like a circuit fault might be exactly what the sequence was supposed to produce. I did not yet have enough cases to know.
The second and third experiments used the same three runes in the same sequence, varying only the channel dimensions. The logic was direct: if aether moved through the channels as current moves through a conductor, then the crosssectional area of the channel ought to affect throughput the way conductor width affected resistance. A wider channel should pass more, a narrower one less. I wanted to test that before changing anything else, because if the channel geometry had no effect I needed to know it before spending time on sequences that assumed it did.
For the second board I deepened the channel cuts by approximately half, pressing the knife blade further into the grain on each pass, running the tip along the scored line twice. The resulting cuts were noticeably wider and deeper than those on the first board, the edges less clean where the wood had pulled slightly at the second pass. I activated the Fehu rune the same way, thumb pressed flat against it, and waited.
The thermal effect returned. Kenaz again, same localized warmth in the carving. But it ran longer I counted seven or eight seconds before it began to fade, and the warmth itself reached further into the surrounding grain, the diffusion radius wider than on the first board. I checked the Thurisaz position carefully. Still nothing. The increased channel depth had changed the duration and apparent intensity of the middlerune effect without producing any response at the terminal end.
I noted that and moved to the third board, where I cut the channels to roughly half the original depth, light scratches that barely broke the surface of the wood, shallow enough that the grain showed through the cut if I tilted the board against the window. When I activated the Fehu rune, there was a pause longer than on either previous board before I felt anything, and what I felt was not warmth at Kenaz but a brief, faint vibration less than a second somewhere in the middle section of the board, not localized enough to say exactly where, and then a sound I was not prepared for.
The board cracked along the channel line with a report like a green stick snapping, sharp in the quiet room, and a thin line of smoke lifted from the split. I pulled my hand back. The crack ran straight from the Kenaz position toward the Thurisaz terminal, following the grain exactly, the wood at the origin several shades darker than the surrounding pine, a dry charred color concentrated in the carved symbol and fading along the channel toward the terminal end. The Thurisaz rune was intact. The channel between the two was not.
I turned the cracked board over, checked the underside, and set it down again. The charring was shallow it had not burned through the wood, just discolored it. But the crack was structural. The board would not hold another activation.
I wrote: Third board. Shallow channel, reduced depth approximately half of first baseline. Extended pause before response. Brief vibratory sensation followed immediately by structural failure crack along channel line from Kenaz toward Thurisaz. Charring at origin. No observed thermal buildup prior to failure, unlike boards one and two. Then: Hypothesis shallow channel acts as highresistance element in series. Same power input, higher resistance, increased heat dissipation at restriction point, catastrophic failure at weakest grain. Too much energy, insufficient path to carry it. The analogy held in the direction I had not tested first.
I lined up all three boards on the desk first intact, second intact, third split and looked at them for a while. Three different channel depths, three different outcomes: incomplete sequence, extended middlerune effect, structural failure at the restriction. The middle value had produced the strongest sustained response without damage, which meant there was a working range somewhere and I had bracketed it on one side.
The crossedline failure came on a Thursday afternoon, working from a design I had spent the previous evening drawing out on a scrap of parchment before committing anything to wood.
What I wanted was a dualinput structure two power runes feeding simultaneously into a logic rune, and from there to a single output. The text described Concordia as an AND condition: output occurred only when both inputs were present. In circuit terms that was a series gate, two switches in series, both required to be closed before current reached the load. I wanted to see whether the runic implementation actually behaved that way, or whether the text's description was theoretical and the physical result was something different.
The layout problem was purely geometric. Routing two separate input channels to a single junction meant the lines had to converge, and on a flat piece of wood the two natural approaches were to run them parallel and let them meet at the junction from the same side, or to cross them somewhere in the middle and let them arrive from opposing directions. I tried the crossing first, reasoning that I should know what the failure mode was before I tried the working arrangement, and failure modes were more informative than successes at this stage.
The carving took longer than the singlesequence boards. I had to place three runes Fehu on the upper left, Uruz on the lower left, Concordia at center, Kenaz at the right and route two channels that crossed at approximately fortyfive degrees somewhere between the inputs and the junction. The crossing point itself I carved through cleanly, one line over the other, no break in either cut. The channels looked correct. The runes were as neat as I could make them with a kitchen knife on a roughcut scrap.
Activating two runes simultaneously required both hands, which produced an awkward grip left thumb on Fehu, right thumb bent backward to reach Uruz at the lower position, the board balanced against the desk edge. I pressed both at the same moment and felt the board respond immediately.
Not warmth. Vibration. A low, rapid tremor that ran through the wood and into both thumbs simultaneously, the kind of sensation that made the tendons in my hands want to retract, lasting no more than two seconds before it stopped. Both input runes went inert under my thumbs. The Concordia junction showed nothing. The Kenaz at the output showed nothing.
I held the position for a few seconds longer, in case the effect had simply been brief, then released and pressed the Fehu rune alone. The familiar Kenaz warmth returned short, localized, the same as the baseline first board. The system had not been destroyed by the crossing failure. I pressed Uruz alone. Nothing: Uruz was classified as an amplifier in the text, requiring a source signal before it contributed anything, which meant it was correctly inert without Fehu providing input. Both inputs individually still functioned as expected. Only the combination, routed through the crossing, had produced the failure.
I set the board down and drew the crossing in my notebook, marking the failure point at the intersection. A short circuit in the electrical analogy two conductors touching where they should not, the potential difference between them driving current across the junction rather than along the intended path, both pathways collapsing into the crossing rather than continuing to the load. The channels were not insulated from each other the way wire insulation kept copper conductors from touching. Any intersection was a junction. Any junction not explicitly defined as one in the runic grammar was a leak.
I carved the arrangement again on a fresh board, this time routing both inputs around to the same side of the Concordia rune, the two channels running parallel for the last inch before reaching the junction from the left rather than converging from opposite directions. The carving took longer and the rune positions were less even than on the first attempt, Uruz sitting lower on the board than I had intended, but the channels did not cross anywhere.
I braced the board against the desk and activated both inputs simultaneously. This time the Kenaz output responded a clear thermal effect at the terminal rune, noticeably warmer and fasterrising than anything a single input had produced, the heat spreading into the grain around the carving within two or three seconds and holding there for six or seven before it faded. No vibration. No crack. The junction had worked, and it had worked exactly as the text described: two conditions, one output, both required.
I wrote: Intersecting channels create failure energy collapses at crossing rather than continuing. Both input paths destroyed simultaneously at junction. Effect consistent with electrical short: unintended connection between two conductors at crossing point. Parallel routing resolves this. Noncrossing convergence allows proper junction behavior. Concordia gate confirmed functional when channels do not intersect.
Then, after a pause: The implication is that channel layout is as critical as rune selection. A correctly chosen sequence with poor routing fails the same as an incorrect sequence. The inscription is not just symbolic it is structural.
Consistency was the problem I could not solve.
I ran the working dualinput arrangement four more times over the following days, carving new boards as carefully as I could, trying to hold every variable I had access to constant: the same rune order, the same approximate channel depth and width, the same pine scraps from the same bundle, the same activation pressure and duration. I wanted to know whether the first successful result was repeatable, which was the most basic requirement for anything I could eventually call a principle.
Two of the four attempts produced the expected output Kenaz warmth at the terminal end, consistent with the first successful board. On the third attempt, the heat response appeared at the Concordia junction rather than the terminal rune, which meant the signal was stopping at the logic point and not propagating through to the output, a result I had not seen in any previous test and could not account for. The junction was not in the right position to be a load. The text did not classify Concordia as an energyabsorbing rune. There was no reason, by anything I currently understood, why it should behave as a terminal point.
The fourth board produced nothing. I activated both inputs, held the position for a slow count of fifteen, released, checked the terminal rune by touch, reactivated, waited again. Nothing from the junction. Nothing from the output. Nothing from the input runes beyond the slight roughness of carved wood under my thumbs. I checked the channels for cracks or splits. The board was intact. I compared it against the two that had worked, holding them side by side at the window where the gray light was strongest.
The rune forms looked the same. The channel depths appeared consistent. The routing was parallel on both boards, no crossing. The pine grain was similar I could not rule out subtle differences in density or moisture, but nothing was visible. I spent the better part of an hour working through the comparison, pressing the nonfunctional board against the desk edge the same way I had held the working ones, varying the activation point slightly in case I had been pressing against the wrong part of the carved symbol, trying one input alone to confirm the input runes were at least partially live.
The board remained inert.
I wrote: Unknown variable present. Four attempts at identical arrangement: two successes, one junctiontermination anomaly, one complete failure. No physical difference between failure board and success boards detectable by eye or touch. Then I sat looking at that for a while.
The unknown variable had several possible addresses. It might be the material grain density varied within a single piece of pine, and if aether conductivity tracked some property of the wood's structure that I could not see, two boards from the same bundle might have meaningfully different conductivities in the carved region. It might be the carving itself a difference in edge quality, in how cleanly the knife had cut rather than torn the fibers at some critical point in the channel, something too fine for my eyes to register but sufficient to affect flow. It might be the activation: intent, the text mentioned in several places, was not merely metaphorical. The runic principles stated that magical expression was modulated by the resonance pattern of the acting soul, which meant my mental state during activation was potentially a variable and not one I could hold constant by deciding to, any more than I could hold my pulse constant by deciding to.
Or it was none of those things and the system had properties I had not yet found.
"You cannot test what you cannot see," I said to the board, and meant it two ways: I could not see the magic moving through the channels, and I could not see whatever internal property of the wood was accepting or rejecting it. I needed an instrument. I did not have one. Both of those facts were going to remain true for some time.
The spacing tests ran over three separate sessions, the results inconsistent enough that I was not certain what conclusion to draw from them by the end.
The setup was a single flat board, wide enough to accommodate a row of runes placed at measured intervals. I scratched shallow channels between them and placed pairs of identical simple runes Fehu at the left edge, and a sequence of Kenaz symbols at oneinch, twoinch, and fourinch removes from the source, no branching, the boards cut long rather than square so the distances were physically meaningful rather than approximated.
The reasoning was this: if aether moved through a channel the way current moves through a conductor, it should attenuate over distance. The further from the source, the weaker the signal at any given receiving rune, assuming no amplification in the path. What I wanted to know was whether the attenuation was gradual and proportional, sharp and thresholdbased, or absent entirely whether distance mattered at all.
First run: I activated the Fehu rune and held it. The nearest Kenaz, one inch distant, warmed in approximately the same duration I had come to expect from closerange singlesequence boards. The Kenaz at two inches showed a faint warmth when I pressed my fingertip against it less than the nearest, noticeably so, detectable but not certain enough to record as a clear result. The Kenaz at four inches showed nothing. I held the activation for a longer count than usual, thirty seconds rather than the ten I had been using as a standard, and pressed hard against the fourinch position. Still nothing.
Second run, same board: The nearest Kenaz warmed as before. The twoinch position produced what felt like a slight elevation in temperature I genuinely could not say whether it was the rune or the warmth of my own fingertip after pressing against the nearby position, which was a confounding factor I had not considered before the run and could not remove from the result after. The fourinch position again showed nothing.
Third run, new board with fresh cuts: The nearest Kenaz warmed. The twoinch position showed nothing. The fourinch position showed nothing.
I wrote: Distance appears to limit effective range of magical influence. Nearrange response consistent. Midrange response inconsistent across runs possibly real but below reliable detection threshold. Farrange response absent in all three runs. Then: Confounding factor in midrange readings cannot rule out thermal contamination from finger contact with adjacent position. Instrument problem again.
The fourinch result was the only clean datum. Whatever the range limit was, it sat somewhere between two and four inches under these conditions, in unsupported pine, with a single Fehu source and no amplification rune in the chain. That told me the limit existed. It told me nothing about where it fell or why, or whether a stronger source rune or an Uruz amplifier in the channel would extend it. Those were further experiments, and I added them to the list.
One arrangement worked once in a way I could not account for, and never again.
It was a circular design, a closed ring carved at the outer edge of a roughly square board, four runes equally spaced around the circumference with channels connecting each to the next in sequence. I had built it as a test of the Third Principle in its simplest form what did containment alone do, without a deliberately ordered sequence? The text's Ordered Flow principle required a directional path from source to result. A ring had no defined beginning and no defined end. If I put a power rune anywhere in a closed loop, which direction did the current flow? Both directions simultaneously? Neither? I wanted to observe a containment failure before I attempted a properly contained design, and I expected the ring to produce either nothing or something chaotic.
The runes I chose were Fehu once, to provide a source, and Jera at each of the remaining three positions. Jera the text associated with gradual energy accumulation a cycle rune, the year turning, energy accruing over time rather than discharging in a burst. If any rune was suited to a nondirectional arrangement it was probably that one, since its function was cyclical rather than sequential. That was the logic. I was not confident it was correct logic.
I carved the ring carefully, the outer circle cut with the knife tip held at a consistent angle, the four symbols placed at what I estimated to be equal intervals. The channels connecting them followed the arc of the circle rather than running straight, which meant each channel was slightly curved, which I did not know how to account for in terms of conductivity. Another unknown.
I activated the Fehu rune with my thumb and waited for either the shortcircuit vibration I had learned to recognize or the localized warmth of a working singlesequence board. What I got was neither. Several minutes passed. I held the activation for a full minute before releasing, checked each rune position individually, found nothing, and set the board on the desk to write down the result as a failure.
It was when I reached for the notebook that I noticed the board was warm. Not at any one rune. The warmth was distributed across the whole piece, low and even, the way a stone takes on warmth from a hearth not hot, not localized, just the temperature of something that had been sitting near a fire for a quarterhour. I picked it up and turned it over. The back was warm too. Whatever it was, it had moved through the full thickness of the wood rather than staying on the carved surface.
I set it down and did not touch it for five minutes by my reckoning, then picked it up again. Warmer. Not dramatically there was no rapid accumulation, no sense that it was building toward something but measurably warmer than before by the test of my palm.
I put the board in my coat pocket and wore it through the rest of the afternoon: the walk back to the dormitory, the hour before supper, supper itself, and an hour of reading in the common room afterward, the board against my ribs through my coat. When I took it out in the dormitory after Thomas was asleep, it was cool. Exactly as warm as the other scraps of pine in my coat pocket, which was to say, about as warm as the room.
The runic text stated that magical constructs decay toward ambient equilibrium without active sustaining which matched what I had observed, since I had not been maintaining any activation after leaving the classroom. But the decay should have been gradual, and I had felt the board through my coat at no point during the intervening hours, which meant I had no idea when it had cooled or whether the cooling had been gradual or sudden. The warmth had been there when I put it in my pocket and was not there when I took it out. Everything in between was unobserved.
I wrote: Possible energy accumulation in closedring arrangement. Distributed warmth across full board, not localized to any rune position. Temperature appeared to increase over fiveminute interval after initial activation. Duration of persistence unknown board found cool after several hours, no intermediate readings taken. Mechanism unclear. No repetition of effect on subsequent tests.
I ran the same arrangement five more times over the following days, using boards I carved as carefully as I could manage. None of them produced any effect. I pressed the Fehu rune and waited, repeated the fiveminute delay, checked the board temperature by palm, held activations for varying durations, tried it in the classroom cold and tried it in the common room warm. Inert. Every one of them inert.
I compared the original board against the new ones and found nothing to account for the difference. The ring proportions were consistent. The rune forms were consistent. The channel depths were as close as I could cut them. The wood was from the same bundle.
I added it to the pile of unknowns and did not look at it again for a week, because looking at it was not going to produce more data.
My notes by the third week were a fair picture of what the work had run into. I went through them one evening and found the following words recurring across every page: warm, warmer, faint, stronger, slight, brief, mild. Every effect I had recorded was described with language that a thermometer would have made obsolete in thirty seconds. I had no thermometer. I had my hand and my eyes and whatever sense of duration a person develops from paying attention to time without a clock, which was not a reliable instrument under any conditions and was especially unreliable when one hand was occupied pressing against a piece of wood and the other was poised over a notebook.
I wrote at the bottom of a page, in larger letters than usual: If magic follows quantifiable rules, then the rules cannot be confirmed without measurement. Currently have no method of measurement beyond qualitative description. This is the limiting factor on all further work.
Then, because it needed saying to the empty room: "You cannot do real engineering without a way to measure."
The echo was not satisfying, but the statement was correct. Everything I had established to this point was pattern recognition. The intersecting channels seemed to interfere. The shallow channels seemed to increase local heat dissipation to failure. The dualinput arrangement seemed to amplify output. The distance seemed to attenuate signal. None of it was more than seemed. I had no numbers, only approximate descriptions, and a theory built on approximate descriptions was a guess with better presentation.
What a working magical measuring instrument would require was a longer problem than I could solve in an evening. The ideal would be something that responded to aether flow in a quantifiable way some material or arrangement that changed state predictably and proportionally with the energy passing through it, giving me a reading I could compare between experiments. I did not currently know whether any such material existed, or whether the runic system included any element suited to that purpose, or whether the entire concept was inapplicable to magic in a way I had not yet discovered. I added it to the list of things to think about later and closed the notebook.
Margaret came by the classroom once, on a Wednesday afternoon, carrying her Potions notes and wearing the expression she used when she had already decided on something and was waiting to see if she needed to explain it. She stood in the doorway for a moment looking at the desk, which was covered in wood shavings and carved pine pieces sorted in a rough grid by outcome working results to the left, structural failures to the right, and the substantial center pile for results I had not yet determined.
"Dost thou want company?" she said.
"I'm fine," I said, which was true.
She looked at the center pile, which was the largest of the three by a margin. "Which are those?"
"Not sure yet."
She nodded, which meant she had understood something I had not intended to communicate, and came in and sat down across from me with her Potions notes. We spent the rest of the afternoon in the same room without speaking much. I carved and recorded. She read and made notes in the margin of her text. When the supper bell rang we gathered our things and went together without discussing whether we would.
The Patronus question started as a side note and became an hour I had not planned to spend.
I had borrowed a different book from the library that same week, a collected account of unusual magical phenomena organized by decade, which I had taken for its chapter on ward structures and the geometry of enclosure in runic work. That chapter turned out to be less detailed than the title suggested a series of case descriptions with no theoretical analysis, the kind of record that is useful to someone already wellversed in the subject and nearly useless to someone trying to derive principles from scratch. I read it twice and marked two passages that might be useful later, then kept reading to see what else the book contained.
Three chapters further in, in a section on Patronus manifestations, there was a case that stopped me entirely.
Most Patronus records described animals. The wolf, the otter, the doe, the hare the overwhelming majority of documented cases produced a creature of some kind, and the established understanding in the literature held that the Patronus took the form of whatever resonated most strongly with the caster's nature or their happiest memory, the two theories never having been cleanly separated from each other in the literature I had read. Various authors accounted for the prevalence of animal forms in various ways: that animals were simpler in nature and therefore closer to pure resonance, that the human soul recognized kinship with the animal world, that the happiest memories of most wizards involved creatures of some kind. None of them addressed directly why animal forms predominated, because none of them appeared to have considered it a question worth addressing. The animal form was treated as the natural result, and natural results do not require explanation by the people who have never seen an exception.
This account described an exception. A wizard, a mercenary in his earlier life who had not discovered his magical nature until his late thirties, produced a Patronus in the form of a soldier. Not an animal that resembled a soldier, not a creature associated with military qualities a soldier, a human figure carrying a sword and a shield, moving, the account said, as a living man moves when he walks. The author, writing in the register of a man collecting curiosities rather than investigating phenomena, noted the unusual form and explained it as follows: the wizard had known nothing of animals in his years of soldiering, and so the Patronus had taken the only form available to it.
I read that explanation twice and disagreed with it twice.
The standard account held that the Patronus reflected the caster's truest nature or their strongest memory. If that was correct, the soldier form reflected the truest thing about this particular wizard that he had spent thirty years fighting, and that thirty years of daily existence as a particular kind of person shaped what a person fundamentally was. The author's explanation assumed the Patronus would have been an animal if only the wizard had known more animals, as though the mechanism were searching through a catalog of options and defaulting to the first available entry. That was not what the theory said. The theory said the form emerged from the caster's resonance, and the caster's resonance was a soldier. The soldier was not a default. The soldier was the answer.
Which meant the animal was not the mechanism's natural output. It was the most common output because most wizards' truest resonance happened to correspond to a creature, or because most wizards' happiest memories involved one. The mechanism itself did not prefer animals. It produced whatever shape the caster's soul was most fully expressed by, and in the overwhelming majority of cases that happened to be an animal, and so everyone had concluded that the mechanism produced animals.
I wrote in the margin: Form follows resonance, not category. The animal is the common result, not the necessary one. The mechanism selects the most resonant shape without regard for what type of thing that shape belongs to. Then: If this is correct, the Patronus form is diagnostic it reveals something about the caster's fundamental nature that the caster may not be able to describe otherwise.
I spent another hour working through the rest of the book looking for other nonanimal cases. The soldier account stood alone. One case in a collected volume spanning four decades of observations, which was almost worse than finding none, because a single case confirmed the possibility without confirming the principle. One case might be a genuine exception. One case might be a misreport. One case told you that something had happened once and nothing else about whether it would happen again or what governed the conditions under which it did.
I put a strip of parchment in the page, set the book with my other library borrowings, and added find more nonanimal Patronus cases to the list of things to pursue when I had time and access to more texts.
The candle on the desk had burned past the halfway mark, the flame low enough that the carved symbols on the nearest board cast small shadows when I held it at the right angle. I had been in the classroom for just over three hours. I gathered my notebook, the runes text, the book with the soldier account, and one small piece of pine with a closed ring on it that had been warm once and was not anymore, and went out.
The corridor outside was cold and quiet, the castle settling into the stillness it found between supper and sleep, the torches in the brackets burned low enough that the light between them was barely sufficient to walk by. I went slowly, not because I needed to think but because I was tired and the dormitory was three floors down and there was no particular reason to hurry.
I had the first rule written clearly enough that I was willing to call it a rule: runes conduct magic, they do not store it or if they stored it, they stored it unreliably in ways I could not yet reproduce, which for practical purposes was the same as not storing it. I had the short circuit observation, which was the most useful single result of the three weeks because it explained a failure mode and suggested a principle. I had the channel geometry results, the inconsistency problem, the distance attenuation I could not measure precisely, and the ring anomaly I could not repeat.
What I did not have was a way to see any of it. The entire body of work described a system I could disturb from the outside and record approximate disturbances in, without any view of what was happening between the activation and the result. If the aether moved through the channels the way I believed it did, I had no instrument that could confirm that movement or contradict it. If the unknown variable was in the wood, I had no way to find it. If the unknown variable was in me in intent, in the particular quality of whatever I was doing when I pressed my thumb against a carved symbol I had even less ability to isolate it, because I could not see myself from the outside any more than I could see the inside of the pine.