Good Style: Economy, Tone, and Audience

• Explain why good scientific style is an engineering constraint on reader effort, not a matter of decoration or personal taste.
• Apply the principle of economy: identifying words, clauses, and sentences that cost the reader effort without adding information.
• Describe what an even, confident tone means in research writing, and the two failure modes, overselling and excessive hedging, it sits between.
• Explain why knowing the intended audience precisely changes concrete writing decisions, not just word choice.

the-shape-of-a-paper-scope-story-and-organization covered getting a paper's large-scale structure right: the right scope, the right order of sections. This concept moves one level down, to the sentence- and paragraph-level choices that determine whether that well-organized structure is actually easy to read once a reader is inside it. Zobel's chapter on "Good Style" and Simon Peyton Jones's independently developed talk agree on something worth stating plainly before any specific technique: style in research writing is not primarily an aesthetic concern. It is a direct, measurable constraint on how much of a reader's limited attention is spent reconstructing what a sentence means, attention that is then unavailable for evaluating whether the sentence's claim is actually true.

Economy: every word earns its place

Economy does not mean short sentences for their own sake; a precise, longer sentence can be more economical than an ambiguous, shorter one that forces a second reading. Economy means every word, clause, and qualifier in a sentence is doing real work, and anything that is not, throat-clearing phrases, redundant restatement, hedges added out of habit rather than genuine uncertainty, is a real cost charged to the reader with no corresponding benefit.

text
Uneconomical:  "It is important to note that, in general, the algorithm
                that we have proposed tends to perform in a manner that
                is faster than the baseline approach in most cases."

Economical:    "The proposed algorithm is faster than the baseline on
                most workloads tested."

Both sentences make the same claim. The second costs the reader noticeably less effort to parse, without losing any actual information the first one contained; the extra length in the first version was entirely padding, not precision.

Tone: between overselling and excessive hedging

An even tone states claims with exactly the confidence the evidence supports, no more and no less. Overselling, claiming a solved problem when the evidence supports only an improvement, or a general result when the evidence supports only a specific case, damages credibility with a skeptical reader the moment the gap between claim and evidence becomes visible, and it often becomes visible quickly, once the reader reaches the actual results section. Excessive hedging has a quieter but real cost: qualifying every sentence past the point the actual uncertainty warrants makes genuine, real uncertainty indistinguishable from reflexive caution, so that a reader can no longer tell which hedges are meaningful.

Voice, balance, and the "upper hand"

Zobel treats a paper's voice, broadly, first person plural ("we present"), passive constructions ("a method is presented"), or a mix, as a real stylistic choice with tradeoffs rather than a fixed convention, though consistency within one paper matters more than which specific choice is made. A related, subtler concern Zobel raises directly is what he calls the "upper hand": the temptation to write in a way that makes a reader feel they must already agree with the author's framing to follow the argument at all, rather than earning agreement through visible evidence. A confident, even tone earns the reader's agreement; a paper that assumes it in advance, through loaded language or dismissive treatment of alternative views, tends to alienate the exact skeptical reader it needs to persuade.

Audience: who, precisely, is this written for

"Know your audience" is common advice stated so generally it is easy to nod along with and then ignore in practice. Zobel's version is more concrete: a paper's intended audience determines specific, checkable decisions, what background can be assumed without explanation, which terms need defining versus which are standard vocabulary in the subfield, and how much motivation a claim needs before a reader accepts it as worth investigating at all. A paper submitted to a specialist workshop can assume far more background than the same result submitted to a broader venue, and writing as though the audience were the broader one, over-explaining standard terms to a specialist audience, costs exactly the same kind of reader effort that economy is meant to eliminate.

Example 1: trimming for economy without losing meaning

Original: "In this paper, what we attempt to do is to present and describe a new approach which we believe has the potential to be able to improve upon existing methods in terms of efficiency." Trimmed: "This paper presents a new approach that improves on existing methods' efficiency." The claim, its subject, and its scope are unchanged; only the padding is gone.

Example 2: matching tone to evidence

A paper's evaluation shows a 12% improvement on 6 of 8 tested workloads, with a small regression on the other two. "Our method solves the performance problem" oversells this result. "Our method improves performance on the majority of workloads tested, with a small regression observed on two workloads involving small input sizes" matches the tone to exactly what the evidence shows, neither overclaiming nor burying the genuine positive result under excessive qualification.

Example 3: adjusting for a specialist audience

A paper on a new Byzantine fault tolerant protocol, submitted to a distributed systems venue whose readers are assumed to already know what Byzantine fault tolerance means, does not need to define the term or motivate why it matters; doing so anyway signals a misjudged audience and costs specialist readers real attention on material they did not need. The same paper adapted for a broader systems venue would need that context restored.

• "Longer, more formal-sounding sentences are more scientific." Economy directly contradicts this: formality that adds no information is a cost to the reader, not a signal of rigor, and Zobel treats padded prose as a stylistic failure regardless of how technical it sounds.
• "Hedging every claim is the safe, honest choice." Excessive hedging is its own failure mode, not a neutral default; it makes genuine uncertainty indistinguishable from reflexive caution, which is a real cost to a reader trying to calibrate how much to trust each claim.
• "Style is subjective, so there's no real standard to meet." Zobel and Peyton Jones's converging, independently-derived guidance on economy, tone, and audience treats style as a set of concrete, evaluable properties, closer to an engineering constraint than a matter of unconstrained personal taste.

Good scientific style reduces to a small number of concrete, learnable properties rather than untaught taste: economy, ensuring every word in a sentence earns its place rather than costing the reader effort for no informational gain; an even, confident tone that matches the actual strength of the evidence, avoiding both overselling and excessive hedging; and a precise, specific sense of the intended audience, which changes real decisions about what can be assumed and what needs explaining. Together these determine how much of a reader's limited attention is spent understanding a sentence versus evaluating whether its claim is true.