#14 Volume biology
Following up on the promise of streaming our premier efforts in learning Finnish like a machine, and predicated on three months of rewarding toil, we claim to be entering a phase that we understand as fine-tuning (in opposition to the hard-core training that preceded, where the terminology borrows from dernier cri engineering). How do you know? Since we craft the lexicon (its new usage in our field), we have full license for its definition and, with that, can hardly be proven wrong. In broad brushstrokes, fine-tuning begins, in our worldview, when recourse to translation (the prompts of learning like a machine) is generally no longer needful for the reasonable-speed reading of large swathes of texts of various sorts. As we can see, it’s a loosey-goosey definition. It entails four arguments so contentious that speaking of a transition (from training to fine-tuning) rather than of a clear cut dichotomy might seem more adequate. (We have observed previously that the machine itself experiences a much clearer boundary: that’s core engineering feeling happy with the model and handing it over to the beta team.) Nevertheless, we feel there is a tipping point. A day, almost, when we find ourselves with a book but no translator at hand, and we take a look, squint a bit, rub our eyes, and yes, we can see! The gaze criteria extending that metaphor might be the most discriminating. Fluency and clarity seem indeed precise counterparts: an understanding that is sufficiently comprehensive and instant so that even a casual glance can catch bits of meaning, just as in the depths of limpid water our eyes can follow the movements of coral and call the show reality. A day of miracles.
For our most quantitative readers, who might not be content with such a parable, we can also try to specify the fuzzy terms more tightly. Recourse to a translation is generally no longer needful: perhaps we look up something once per spread in-douze (the pocket book), once per page in-octavo écu (the university textbook), three or four times per page of newsprint. That’s on average. Perhaps things get very blurred all at once and we must sieve a whole quarter. Here, must is an embedded fuzzy modal verb that invites clarification. It can be blurry without the overall comprehension of the chapter being compromised. Yet, you felt some nuance flying over your head. In fine-tuning, and provided we do have a translator at hand for sparse look-up, our tolerance for nuances escaping is low. We are in for transparency. If it resists, maybe we read twice. Maybe we fell asleep a little, we stretch our back, stand tall, and backtrack to the penultimate sentence. Now there are short and longer look-ups. A gap is left open that we must fill when a word is missing in our lexicon and we cannot, in spite of a reasonable time spent trying, infer it from the context. Or perhaps an entire clause just won’t make sense. We have glued together pieces based on what we think the syntax prescribes, but the resulting collage doesn’t gel into the broader text. We type into our interactive translator the most obscure bits, or the whole thing, and perhaps we play around a bit, translate back and forth, as we did when we were in a training mood. Otherwise, a less absolute, more conscious must might encourage us, once in a while, to check a word or turn of phrase we’re fairly sure about: to hammer home the point. Texts are of various sorts when they span different registers, topics, styles or epochs. This might be the most relaxed of the loose terms: we already claim to have reached fine-tuning when we can proceed through some (wider) class of texts that are (complex enough yet) most transparent to us (back to that in a minute). While speed cannot be expected to fully match the native one, it’s not extremely far behind. There must be nothing in the way: the right image is this time that of the mountain torrent that trips delightfully from rock to rock unimpeded. From all the factors specified above (speed, technical breaks along the way) derives the possibility of reading large swathes of text. To read ad libitum, a misnomer, for all good things (as they currently stand) must end: to read until the book falls off our hands.
Based on its very specification, fine-tuning invites, allows for, and necessitates volume reading. Why volume is requisite, and how, ultimately, fine-tuning works, is what we will attempt to clarify in what follows. Attempt, without any guarantee of a full elucidation that we haven’t waited for in order to proceed. The cool thing about pragmatic sciences is that, albeit perhaps not an explanation, it works suffices as evidence.
Why strive to explain, then? Following such a thread, you end up digging up a kilometric underworld of cables. Why do you need explanations at all? When do you deem one to be explanatory enough? For the time being, let’s simply suggest two motives. Perhaps we want to disseminate our intuitions. To do so requires us to express ourselves, that is, to place words onto things, which, on the face of it, may well be what explanation is about: our raw instinct is preverbal, its verbalization a making explicit that stands in for a preliminary elucidation. For instance, the ingenuously descriptive the Earth orbits the Sun already entails a clarification of how that portion of the universe works. In our case, the bare instruction read plenty is implicitly explaining how you’ll make progress. One might argue that it nonetheless falls short of answering why, which, beyond depicting, seems to need words to meddle somehow in the causal relationships at stake. In the case that interests us, something along the lines of reading plenty causes X (that causes Y etc.) that causes progress at the end of which the fog dissipates and you sigh: Ah, that’s why. Once grounded, to a depth at the discretion of the teacher, the lesson might spread more convincingly. In addition, deep justification helps tune or revolutionizes practice. Variolation is documented in China from the 16th century, Edward Jenner inoculates patients with cowpox in late 18th century England, fully blind to the structure of antibodies that emerges no sooner than around 1960. Yet, the belated molecular specifics don’t seem superfluous: coming from the theory side, they corroborate the naturalness of what had been observed experimentally. They account for variability and failures, fence off the field where safe extrapolation pastures, support precision design, public health campaigns and control at scale. We cherish the hope that epistemic roots growing under our serendipitous reading plenty can likewise make it stronger. Contrary to those of trees, which stretch unseen underneath the fern lattice before the trunk arises, such roots cannot predate laborious instincts that emerged through frantic trial-and-error, just as astonishment at the efficacy of vaccination was the trigger requisite for immunology research. Ultimately, that’s the teacher in us, the self-coach, who first pockets the profits of reverse-engineering. There might be, for instance, a special link that weighs most in the magic, and that, once discovered, can be taken advantage of. Wherein being simultaneously the lab rat, the alchemist and the control student must be acknowledged to be a privileged superposition. Hypotheses pop up by doing, get confirmed by trying, the theory builds up on the move while we learn. Yet, the very same effort opens the door to servicing the wider public: it will extirpate the special case from its happenstance, without which the knack can’t aspire to universality. (It will incidentally help disarm the doomsayers who might cling to the convenient belief that one « must have a head for Finnish » (something hard, written in the genes).)
Can you tell us in the end why volume reading works? Of course, that’s a gigantic endeavor which we could keep working on in perpetuity. But we can, as of now, intuit a proof whose key argument sketches out as follows: going through text lines aplenty exposes you to pure syntax, which instills in you the spontaneous music of the language. Within this purely formal latticework, the lexicon intrudes, which the text’s continuous cohesion hammers home.
Isn’t volume exposure the crux of learning languages like a machine in general? You are a disciple. That’s nearly right, and we need to specify what it means to volume-read in the fine-tuning phase, which is what we’re talking about here. Reading is not matching stricto sensu. It’s matching to meaning on the fly, to the virtual native prompt that gets woven as your cursor window marches. Volume reading being specifically the unhindered (overwhelmingly) monolingual traversing of wide expanses of literature, one body of work at a time, its efficacy deserves a scientific explanation of its own. While training by matching invites lengthy pondering and grammar textbook strolls, the slow peering at the forms, through a fog that scarce rushes of light interrupt, fine-tuning crushes the pages like a barrel organ. Once cruising speed is reached, the engine no longer even has time to spit out the script: we remain tight-lipped, it‘s understanding, subliminally, within ourselves.
We can’t fine-tune at any stage on anything. A systematic order emerges. We don’t start by volume-reading novels. Literature is a hydra that speaks as many languages as there are authors. Since imagination is free-flowing, no special backbone can be suspected: an invertebrate hydra into the bargain. The diametric opposite has our preference in this case: the college textbook in the special sciences. And here we are, back from the woods, volume-reading in a basement tome two of the canonical Finnish college textbook in immunology. (Heikkinen, 2024)
We’ve reached page two hundred and the end of section one, Yleisimmunologia: a good time to stretch a bit and reflect on our argument. Scientific literature intended for pedagogy features the epitome of explanatory discourse. It’s just a big answer to how does it work, and, digging at different scales, why does it work the way it does. A quick-and-dirty demonstration could be content with: we volume-read biology for it’s a Cartesian cut into well-known phenomena, odds are high that we will understand, that native clues will flash before our eyes, just lengthening our training at bilingual matching. There is some fair share of truth in that: not only did we get acquainted earlier with the subject, but its most notable encounter happened in Estonian, a language cousin to Finnish that partly shares with it stems and airs, especially in the specialized lexicon. (The Estonian counterpart to canonical scholarly immunology was our first fine-tuning book while learning Estonian like a machine. We are trying to invoke the same spell, feeling like we have our hand on a fetish.) In sum, that’s working for that’s slightly pre-chewed, with better prospects of understanding. Still, there is more to the matter, we think.
Something that resides, as we alluded to, in the purpose being explanatory. What is so special about an explanatory tone? Remarkably, it doesn’t leave much room for surprise, which we illustrate with a quote (its translation, we’ll soon look at the Finnish original too) from chapter nine on cell-mediated immunity.
The most important task of activated helper T cells in immune defense is …
A task. We expect an action of some sort, tangentially a change of state, a flick to a domino in a sequence.
… to direct and assist the activation and function of other cells. …
It’s nearly inevitable that what comes next will zoom into the specifics of that task. It might unveil the micro-steps it takes to direct and assist, the modalities or conditions of the activation and function, the nature of the other cells. And indeed
… The assistance is partly dependent on the binding of molecules on the surface of …
It brings forth what’s decisive in the unfolding of the task. Not at random, but conditioned by something belonging to… necessarily to either the agent, namely the activated helper T cells, or the object hanging at the other end of the causal link, the other cells.
… of molecules on the surface of the helper T cell to the corresponding …
If that’s corresponding, that’s of the same brew.
… the corresponding molecule of the target …
Cell. What comes next? Perhaps it is now the turn of these molecules to come under scrutiny. Or a bracket closes, and another concept from the first sentence unfolds, for instance, activation or function. It’s the former:
… An important molecular pair here is the CD40L of the helper T cell and …
A pair. Of molecules. One sits on the agent. We expect, with probability 0.98, another bookkeeper acronym standing for a molecule that sits on some target cells.
… and the CD40 found on the surface of for example macrophages and B cells …
Sentence three could not naturally be glued right after sentence one:
The most important task of activated helper T cells in immune defense is to direct and assist the activation and function of other cells. An important molecular pair here is the CD40L of the helper T cell and the CD40 found on the surface of …
A link is missing: that of the cells kissing through their surface molecules. We do roughly get it, but it requires our imagination, maybe helped by remnants of high school physiology, to bridge the gap. Sentence two could have brought another argument of the premise into the spotlight, such as the direction or the activation, but not suddenly branched into an isolated consideration on macrophages or B cells. Now imagine a witch (a hex, to borrow from Kyburg) had sneaked in! Witched activated helper T cells have surface CD40L molecules that long for mooring other cells’ CD40 ones. Formally, the proposition remains true, while the goodness of the explanation sinks: it’s sufficient for the cells to be activated, and that they are witched into the bargain is irrelevant. On top of that, in the context of twenty-first century science, witching is known to be an imposture, and the explanation is simply dismissed as heretic.
In other words, we seem to be instinctively equipped with fairly specific expectations. What comes next, for the explanation to be relevant and good, will probably answer a why-question on one or several of the components in the thing to be explained, or dig down a hole already primed. As we also suggested in passing, that answer will prune irrelevancy. It will preferably be decisive, as opposed to secondary or contingent: it will make a difference in the train of events leading to the matter under consideration. That chain we expect to be somehow causal and finally, we let high-school biology chime in, that is, known facts or consensus, something quite universal, akin to laws. We here let the interested reader dig into the retinue of historical theories of scientific explanation that our above illustration ultimately hovers around, and must credit Woodward et al. and Kitcher for having provided us with good reads (Woodward, 2025; Kitcher, 1989). Kitcher prominently caught our attention. While previous theories of explanation, from derivation models underpinned by laws of nature (Hempel) to causal approaches, evaluate the « explanatory worth » of candidate statements individually, Kitcher’s unificationism strives to set conditions on a set of argument patterns (or « explanatory store » E(K), K being the set of statements endorsed by science at a particular time) from which phenomena are derived. An explanation’s worth is measured by that of the « explanatory store » that the argument pattern it abides by belongs to. Kitcher’s argument patterns are, more precisely, the triple of a schematic argument, itself a sequence of schematic sentences (the logical backbone of a sentence whose lexical arguments have been replaced by dummy letters); filling instructions constraining how to fill the dummy letters; and a classification that characterizes the inferential unfolding of the schematic argument (which sentences are inferred from which, according to which inference rules, etc.). The ideal « explanatory store » is then, in a three-way trade-off, the minimal set of argument patterns of maximal stringency from which a maximal number of phenomena can be derived. It is, in other words, the « explanatory store » that best unifies K.
And here, we can’t help but think of the Smurfs. Interestingly, their everyday language tends indeed to indulge in Kitcher’s gymnastic, freely substituting lexical items that are secondary to the sentence’s logic with smurf and its derivatives.
The most important task of smurfs in smurf is to direct and assist the activation and function of other smurfs.
In Smurf Versus Smurf, two Smurf subcultures clash in a conflict of the grammarians. One of them claims that the proper use of the swiss-knife word smurf is exemplified by bottle smurfer, the other that the correct way is smurf opener. With such dialectal variations in mind we could also rewrite the same into
The most important task of activated helper T cells in immune defence is to smurf and smurf the smurf-ing and smurf-ing of other cells.
or even smurf it all
The most important task of smurfs in smurf is to smurf and smurf the smurf-ing and smurf-ing of other smurf.
In Kitcher’s worldview, explanatory productions of the best quality expose us at length to a small set of smurfed arguments that run in a loop to produce plenty of conclusions that hold up. Additionally, the smurfed items in there can’t be stuffed back at random, but are stringently constrained. An obvious objection points to the lack of a general theory for assessing the trade-off optimality of those qualities, which Kitcher sidesteps by proposing that an « explanatory set »’s goodness will be clear enough. We can, in any case, be content with an immunology textbook that’s explanatory enough and assume it will do the job of flashing us with pure syntax (the logical ossature in linguistic utterances, left after those were stripped of their stuffing marrow).
We recall our excerpt (reading on a bit)
The most important task of activated helper T cells in immune defense is to direct and assist the activation and function of other cells. The assistance is partly dependent on the binding of molecules on the surface of the helper T cell to the corresponding molecule of the target cell. An important molecular pair here is the CD40L of the helper T cell and the CD40 found on the surface of for example macrophages and B cells, but many other molecules also modify the signalling between cells. Helper T cells are also effective producers of cytokines. As a result of the differentiation of function following activation, different helper T cells produce different cytokines, which have different targets of effect.
which we’ll now smurf the Kitcher way, swapping the marrow with algebraic unknowns.
The most important task of A in B is to C and D the E and F of other G. The H is partly dependent on the I of J on the K of the L to the corresponding J of the M. …
Just as in the Smurfs, there is ground for quarrel over the rules governing the algebraization. First, as earlier with smurfing, one can be more or less strict about what is skeleton and what is marrow. But there is more. Our dummy letters are not independent of each other, but subject to constraints: in line with Kitcher’s filling instructions.
A → activated helper T cells
B → immune defense
C → direct (the base form of the verb)
D → assist
E → activation (this is a verbal noun)
F → function (that too, arguably)
G → cells
A first kind of restrictions pertains to the part-of-speech, noun or verb. There are a lot more, we won’t aim for exhaustiveness but just name a few.
A → a noun, a potential agent (something that can have a task)
B → something thematically related with A, and, in fact, something encompassing A and its task
C, D → bare infinitives, things that A can do and that belong to A’s most important task
Here we suggest that the more we know about A, the more stringent the restrictions are. If you can’t rule out that activated helper T cells are akin to gymnasts, you can’t exclude either that their most important task is to thread the needle or go up on pointe. (Stringency would be, with that, in the eye of the parser, where we intuit Kitcher’s is congruent with that of the foremost scientist of the time.)
G → G is a superset of A
Perhaps your algebra should factor such a dependency in. We could (shifting all symbols) represent cells by A and activated helper T cells by B-A. We could also encode the verb part-of-speech somehow, for instance with
do-E → assist
but we’ll simply replace by letters or their starred version for indicating part-of-speech variations of the same brew (e.g. E for assist and E* for assistance).
The most important task of B-A in C is to D and E the F and G of other A. The E* is partly dependent on the H of I on the J of the B-A to the corresponding I of the target A.
In the same vein, we get tentatively to
An important I* pair here is the K of the B-A and the L found on the M of for example N and O-A, but many other I also modify the P* between A. B-A are also effective Q* of R. As a result of the S of G following F, different B-A P different R, which have different targets of effect.
You might object we lost sight of the Finnish, and that’s a good time to pull it back in.
Aktivoituneiden auttaja-T-solujen tärkein tehtävä immunopuolustuksessa on ohjata ja auttaa muiden solujen aktivaatioita ja toimintaa. Apu on osin riippuvaista auttaja-T-solun pinnalla olevien molekyylien sitoutumisesta vaikutuksen kohteena olevan solun vastinmolekyyliin. Tässä tärkeä molekyylipari on auttaja-T-solun CD40L ja esim. makrofagien ja B-solujen pinnalla oleva CD40, mutta monet muutkin molekyylit muokkaavat solujen välistä signalointia. Auttaja-T-solut ovat myös tehokkaita sytokiinien tuottajia. Aktivaatiota seuraavan toiminnan erilaistumisen johdosta eri auttaja-T-solut tuottavat eri sytokiineja, joilla on erilaiset vaikutuskohteet.
Let’s rip off the marrow.
B-A tärkein tehtävä C-ssa on D ja E muiden A-n F ja G. E* on osin riippuvaista A-B-n J-lla olevien I-n H-sta vaikutuksen kohteena olevan A-n vastin-I. …
In passing, we are quick to notice that inflecting the algebra is a must: there is logic in the inflectional suffixes the same way there is logic in the English prepositions. Our volume-reading is now a perpetual parsing of backbones that exposes us tirelessly to syntactical patterns (how sentences are built, incl. word order, inflection), the way fundamental relationships are expressed, such as necessity, concession, congruence or causation. Backbones are fleshed with entities that we don’t need to know by their name to march forth, but that aren’t distributed at random, that knowing a bit constrains even more, and that our marching forth will accustom us to. To the point they’ll become second nature, and in turn, catalyze the backbone parsing as we progress. By kneading the bone and the marrow haloing a sempiternal theme, you end up parsing like a butcher’s mincer.
Are you happy with your explanation? We could pass it through the sieve of Kitcher’s theory of explanation, but we must say we are content with intuitions upholstered with expert terminology from a rapid-fire literature review, all the more so since our attention is required somewhere else: towards the very progress of our fine-tuning like a machine. The detour is certainly rich in practical insights, or at least confirms our instinctive ritual. At the start, we shall prefer explanatory texts, whose bare syntax rolls from page to page, and let ourselves be drawn along at full speed by this predictable music that hammers the language within us.
Yet, we must admit we are left hungry for more with regard to the text itself. Immunoloogia, its Estonian counterpart from our previous learning Estonian like a machine, felt like a narrative of better quality: clearer and better organized. (Uibo, 2015) Yet, here as there, a relative degree of mess is to be noticed. Chapters in biology books not being mutually exclusive seems symptomatic of, more than editorial choices, some systemic stumbling block: the living doesn’t split into chapters really, and chopping it neatly is an ambition that’s hard to marshal. No matter how you slice it, all you get is a volume of heated mess whose Brownian motions aren’t to be disentangled, no matter how much you squint. Much happens between the words that the positivist minds populating our time place upon things.
Let’s get back to reading.
References
Heikkinen, T., Järvinen, A., Meri, S., Vapalahti, O., & Vuopio, J. (Toim.). (2024). Immunologia: Mikrobiologia, immunologia ja infektiosairaudet, kirja 2 (2. uud. p.). Kustannus oy Duodecim.
Woodward, J., & Ross, L. (2025). 20th century theories of scientific explanation. In E. N. Zalta & U. Nodelman (Eds.), The Stanford Encyclopedia of Philosophy (Winter 2025 ed.). https://plato.stanford.edu/entries/scientific-explanation-20th/#SRModeLowProbEven
Kitcher, P. (1989). Explanatory unification and the causal structure of the world. In P. Kitcher & W. C. Salmon (Eds.), Scientific explanation (Minnesota Studies in the Philosophy of Science, Vol. 13, pp. 410–505). University of Minnesota Press.
Uibo, R., Kisand, K., Peterson, P., & Reimand, K. (2015). Immunoloogia. Tartu: Tartu Ülikooli Kirjastus.
Pedagogical afterword
The essay has been written without copy-pasting any piece from the text, of which we had only a hard archive copy to be consulted in the library's silent room (one of those where you can't even get in with a loaded fountain pen and the documentalist watches over you). So that's where we read volume biology like a machine, and keyed, word for word, a page-long transcript. Keyed is a misnomer though, as it evokes clerical typewriting. Instead, we read the text conscientiously and transcribed it in full from memory, proceeding a large piece, a sentence maybe at a time. Not the memory of a mechanical sequence of letters, striking the next key as if telling the beads of a rosary of tokens, but that of a sense we would translate back into Finnish, half because we just read the answer key, half through reasoning. Such an exercise was, back in our school days, called self-dictation: a dictation where the school teacher could go on a break, since we were writing down, not what he would have dictated live, but a small piece of primary school literature we had had to learn by heart the previous evening at home. We transcribed the text through self-dictation, and make a point of doing so, with other texts we encounter, as much as possible. Why? For the exact same reason we don't get the parcel box open by copying the six-digit code. Here's how we do it: we purposefully glance at the code briefly, carry on with our life for fifteen minutes, shop at the mall, then retrieve the code from memory, with that, open the door and get the parcel. Are you serious? We want to remain whole.
Obviously, supermarket parking lot micro-exercises are far dwarfed by more ambitious tasks, as is transcribing a Finnish university immunology textbook. Self-dictation thus has a prime place in the panoply of aids in learning languages like a machine. It’s nevertheless one that’s challenging to carry out, for it’s easy to cheat: a glance at the text that’s a little too frequent, and the dictation turns into early-grade word spelling tests. This is still a variant of matching (from meaning to the original text, drawing on memory), but one that calls for a special level of discipline. We illustrate below extremes on the spectrum of rigor that implementation can exhibit.
Stimuloituneiden solujen pinta-antigeenien ja tarttumismolekyylien ilmentyminen lisääntyy, mikä vahvistaa solujen keskinäisia interaktioita.
If we are indeed at the fine-tuning stage, we understand that:
The expression of surface antigens and adhesion molecules on stimulated cells increases, which enhances cell-to-cell interactions.
An extremely poor way to proceed, as you can anticipate, consists in reading the Finnish one word at a time, and indulging in a task that pertains to Taylorism more than linguistics.
(Reading) Stimuloituneiden
(Blank page) Stimulo- … what comes after that, an i?
(Reading) Stimuloituneiden
(Blank page) indeed … Stimuloitun-x (it gets sour in the end, a bad presentiment develops)
(Reading) Stimuloituneiden … we knew it … Stimuloituneiden solujen
(Blank page) Stimuloituneiden solu- … etc.
If that’s the task you’re in for, you can interchangeably learn six-digit codes or the list of flights taking off from Berlin-Tempelhof. Abiding by higher standards, you could split the text in smaller portions that you deem a priori to fit in short-term memory.
(Reading) Stimuloituneiden solujen pinta-antigeenien … that doesn’t quite form a logical unit, but that seems to fit … of surface antigens on stimulated cells
(Blank page) on (of) stimulated cells … Stimuloitun-x solujen … suffix of the plural -i- … suffix of the genitive -den … Stimuloituneiden solujen … surface … pinta- … of surface antigens … pinta-antigeenien …
Typically, we reconstruct this type of situations upside-down (for we do think upside-down, compared to Finnish). Thus, an alternative, and likely, train of thoughts is:
(Blank page) of surface antigens … (leaving a large blank) ___ pinta-antigeenien … on cells … (filling that blank little by little) __ solujen pinta-antigeenien … on stimulated cells … Stimuloituneiden solujen pinta-antigeenien
(Reading) we got that right, now comes … ja tarttumismolekyylien ilmentyminen … and the expression of adhesion molecules …
The issue here is that our left-to-right break-down fails to match the brackets of the syntax. The grammatical subject indeed reads [the expression] [[of surface antigens] and [adhesion molecules]] [on stimulated cells] where the head noun the expression and its adverbial of place on stimulated cells distribute over both surface antigens and adhesion molecules: a possible logical cutting of the Finnish would then be
Stimuloituneiden solujen / pinta-antigeenien ja tarttumismolekyylien / ilmentyminen …
but this time, the memorization challenge deflates. Alternatively, the whole of
Stimuloituneiden solujen pinta-antigeenien ja tarttumismolekyylien ilmentyminen …
might intimidate: you dare not, you’d overflow. We feel that there is a tension between a left-to-right memorization (which is very sensible, since eventually the exercise is not one of translation proper, but one of Finnish dictation) and regard for the integrity of syntactical units (which is pivotal, since whenever memory fails, reasoning takes over that can't make do with incongruous spaghetti). Tangentially, he who chops finely walks in a harness: certain not to fall but predicted not to go too far. With that, here comes the virtuous extreme, the peak, Homeric way to « copy » a Finnish sentence onto a blank page:
(Reading) Stimuloituneiden solujen pinta-antigeenien ja tarttumismolekyylien ilmentyminen lisääntyy, mikä vahvistaa solujen keskinäisia interaktioita.
(Blank page) (in a rush, for the sentence threatens to spill out) ___ , mikä vahvistaa solujen interaktioita … (the most salient portion in mind is quite systematically the last one) … the expression increases … __ ilmentyminen lisääntyy, mikä vahvistaa solujen _ interaktioita … which enhances cell-to-cell interactions … interactions of cells between themselves … keskin- … __ ilmentyminen lisääntyy, mikä vahvistaa solujen keskin-sia interaktioita … (we are no heroes either, and while the partitive seems to prescribe a -sia suffix, we might miss a piece somewhere) … (overflow)
(Reading) Stimuloituneiden solujen pinta-antigeenien ja tarttumismolekyylien ilmentyminen lisääntyy, mikä vahvistaa solujen keskinäisia interaktioita … (perhaps we remember where our blanks are and go thoroughly through the corresponding pieces in a low voice)
(Blank page) __ ilmentyminen lisääntyy, mikä vahvistaa solujen keskinäisia interaktioita … (we are no saints either, maybe we glance at the Finnish and confirm that keskinäisia is right) … expression of surface antigens … pinta-antigeenien ilmentyminen … surface antigens and … something molecules … adhesion molecules … adhesion … tarttu- … tarttuminen? … tarttumismolekyylien ilmentyminen … _ pinta-antigeenien ja tarttumismolekyylien ilmentyminen lisääntyy, mikä vahvistaa solujen keskinäisia interaktioita. (overflow)
(Reading) etc.
We’d much rather make plenty of mistakes, and thus back and forths, in a full-fledged version of the exercise, than excel in a half-baked version. You are the sole judge of the discipline slider, since there’s no feedback loop to reward you with a chocolate or pixelated medal. And yet, the behavioral difference becomes quickly apparent from the results: slow work, fast learning. Albeit with much assistance, the feat starts to activate the muscles of language production, the ones that work when obliterating the « writing » and « speaking » sections in standardized tests.
Lastly, if this observation belongs in the epilogue, that's because it is not the appanage of fine-tuning. We practice it from the very first days of learning a language like a machine. Obviously, the length of the pieces accurately kept at once in memory, and with that, the speed of execution, depend on the progress already made.
Learn Finnish like a machine isn’t run by a billionaire. Help us grow better at our trade and tip us $2.42 today.