學術寫作要訣（1）「Tips to Article-Writers」

mitmgmtfaculty.mit.edu

1 Motivate the paper. The first question you must answer for the reader is why they should read your paper. There is A LOT out there to read and it is very easy to find an excuse not to read a paper. Most people dont even read all the articles published in their fields flagship journals. So if you want your paper to be read, you need to sell the reader on why your paper is so great. The introduction of your paper has to be exciting. It must motivate the reader to keep on reading. They must have the sense that if they keep on reading, there is at least a fair chance that they will learn something new.

intro就是廣告。

當下是拼文案和流量的時代。類似於app想辦法保持住日活躍用戶：只要有流量，就有效益。科技文為了能夠有好的引用，除了科學知識要紮實之外，就是科學家作為作家的基本功：引人入勝。

2、Know your audience. Since different people get excited about different things, you cannot get them motivated unless you know their taste. And different academic communities/journals have very different tastes for what constitutes an interesting question and what constitutes a compelling approach to a question. (My friend and colleague Roberto Fernandez has an excellent framework for thinking about audiences, known widely at Sloan as 「Rows and Columns.」 I will not go into it here, but the basic idea is that social scientific communities are arrayed by two dimensions, where the 「rows」 are 「phenomena」 [e.g., area studies; topics such as entrepreneurship or racial inequality] and the 「columns」 are disciplines or theories. One key lesson is that one typically needs to choose whether one is aiming for a 「row」 audience/journal or a 「column」 audience/journal, and motivate/frame one』s paper accordingly. Trying to motivate both row and column simultaneously usually does not work).

精確定位讀者群。

根據文章「現象」（寬度）「理論」（深度）果決地取捨文章切入點和目標期刊，魚和熊掌不可兼得。結果新穎的文章，現象一端多做拓寬；結果踏實的文章，理論路線多深入。可以那快報和全文類文章做類比。

3、Use substantive motivations, not aesthetic ones. By an aesthetic motivation, I mean that the author is appealing to the readers sense that a certain kind of theory or approach should be preferred regardless of its explanatory power (e.g., we should be avoiding 「economistic」 or 「functionalist」 or 「reductionist」 explanations). Sometimes aesthetic motivations work (for getting a paper accepted), but the contribution tends to be hollow because the end of research (figuring out how the world works) is sacrificed for the means (telling each other how much we like

繡花枕頭爛草包是不行滴。

不能因為某個理論更漂亮就拋開實際價值採用之、鼓吹之，即便我們喜歡漂亮的東西。

4、Always frame around the dependent variable. The dependent variable is a question and the independent variables are answers to a question. So it makes no sense to start with an answer. Rather, start with a question/puzzle! (Note that I don』t mean the literal dependent variable in the analysis in the paper, but the larger process/pattern that it is supposed to represent).

行文應是圍繞一個問題展開，如偵探小說逐步解謎；而不是法庭上先設定有罪或無罪，然後用證據強行證明或駁斥。

5、Frame around a puzzle in the world, not a literature. The only reason

圍繞一個問題/謎團行文，而不是一篇文獻發現沒有人報道就上馬研究。

沒人做過，不代表有價值；具有可探索性以及其附帶因果邏輯的問題，才有價值研究。

intro里引出自己研究內容的用詞要注意了，雖然可能是說同一個研究，但表述為since X has not been studied, we...而不是陳述其涉及到的前因後果，就可能讓人覺得這是一個為了研究而研究的無價值工作。也可以簡單理解為「包裝」。

6、One hypothesis (or a few tightly related hypotheses) is enough. If people remember a paper at all, they will remember it for one idea. So no use trying to stuff a zillion ideas in a paper. A related problem with numerous hypotheses is that it』s never clear what implications the invalidation of any one hypothesis has for the theory. (Note: the organizations community apparently does not agree with me on this one)

一篇文章一個核心假設，足矣。

科學，就是建立和測試假設。但一篇文章中假設太多的話，場面就會混亂。就如同一個人物關係錯綜複雜的故事，往往會失去很多欣賞、支持者。

7、Build up the null hypothesis to be as compelling as possible. A paper will not be interesting unless there is a really compelling null hypothesis. If there is no interesting alternative to the author』s argument, why would anyone care about it? Flogging straw men is both unfair and uninteresting.

零假設要盡量有力有理。

敵人越強大的超級英雄電影，主角也越鮮明；對手都是弱雞，超級英雄就是弱雞英雄，甚至不是英雄。作為導演和編劇，要能夠塑造強大的反派。具體參見有希斯萊傑的Batman。

8、Save the null. Since the null is compelling, it must be right under certain conditions. The authors job is to explain to the reader that s/he was right to believe x about the world, but that since x doesnt hold under certain conditions, s/he should shift to belief x`. This helps the reader feel comfortable

解救零假設。

反派不能太擺設，劇本中要拿出空間給反派，用來展現反派的統治力。讀者看到一個惡統治未來的昏暗無光後，才能自然地理解正義的光明。經過切身的對比，最後選擇才是自然的，不遺憾的。

9、Orient the reader. The reader needs to know at all times how any sentence fits into the narrative arc of the paper. All too often, I read papers where I get lost in the trees and have no sense of the forest. The narrative arc should start with the first paragraph or two where a question/puzzle is framed and lead to the main finding of the paper. Everything else in the paper should be in service of that arc, either by clarifying the question or setting up the answer (including painstakingly dealing with objections). A related tip is:

在問題——答案思路下，引領讀者。

文中句子既然存在，則必要有一個功能，且這個功能只能是服務於提出問題——解決問題這條主線。於此主線無關的內容只會讓讀者茫然不知所措。讀者一迷惑，就給差評。

10、Never write literature reviews. No one likes to read literature reviews. They are borrring. So don』t write them. But that doesn』t mean you should ignore 「the relevant literature.」 To the contrary. You have raised a puzzle about the real world (see tips 3-5). One reason why it is a puzzle is because existing answers are compelling (see point 7), but flawed. So you review the literature not as an end in itself but because you show what is compelling but flawed about existing answers. Any research that does not pertain to that objective can remain unmentioned. (Ok, ok. Some reviewers will demand to see their names or that of their favorite scholars even when their work is essentially irrelevant. And it is usually good to anticipate that. But try to do as little as possible.).

不要寫文獻綜述。

沒人喜歡讀文獻，文獻很無聊；讀者喜歡謎語，也喜歡與謎語相關的提示信息。

寫這個提示信息。

是文字遊戲么？不是。這裡是要強調關聯性的重要性。

沒有關聯性的文獻只是為了引用而引用，為了論述而論述，對解決當下的問題沒有幫助。

引用文獻務求按照尋寶線索的路子去設置。

總結：

Zuckerman這篇文章介紹了幾個核心概念

• 強調故事性 第1、10條
• 強調問題-答案這個核心模塊 第4、5、9條
• 強調故事情節的戲劇性 第 6、7、8條

科技文最難的其實就是在專業知識本身枯燥的無奈下，想通過結構設計讓論文具有一定的可讀性。

基於此，所有核心的建議，或是看似無關的細節，其實都是在為故事性服務。

作為一篇論文的作者，科學家其實也是作家。

用心於此，論文才能不僅僅是一個說明書。

補：

謝謝@蔡先生的提問

null hypothesis」該如何理解呢？它與論文里的那個核心假設區別在哪裡，可否用個簡單的例子說明一下?

首先，原文中並未給出解釋，於是我查了下，null hypothesis是一個既定術語。

這裡我犯了一個錯誤。

我以為這個術語是我理解中的naysayer，即，在一個命題的論述中，自己主動插入一些自己模擬出來的對自己的批評和反駁，從而讓自己的觀點更為有理有據、中立、全面等。有點兒像現在軍事理論發展中越來越重要的朱日和的紅藍軍對抗演戲。這裡換了名字只是不同行業的叫法而已。

顯然，我錯了。

null hypothesis：

零假設不過是我們研究兩個事物之間是否有關的時候我們能做出的最保守的假設。如果我們想知道溫室氣體排放的增加和全球變暖之間的關係，而在研究之前我們沒有做過任何的研究和調查，我們最保險的辦法就是認為這兩者之間沒有關係，這就是零假設。因為在研究之前我們不知道它們之間是什麼樣的關係，所以我們寧可認為它們之間沒有關係。然後通過我們科學的調查研究，再排除那些偶然性的因素之後，我們得到的結論是溫室氣體排放使全球變暖。

今天認真看完了百科的解釋，才發現，兩者的差異。

現在說核心假設和零假設的區別。

• 核心假設的點，在於一篇文章中，只要有一條邏輯線即可：一個重要、靠譜兒的假設，一個驗證，一個驗證結果，往往是假設成立。簡明扼要，容易理解和記憶。
• 零假設的點，在於預處理的過渡價值：在沒有確定證據之前，先搭出一個假設，讓後面的論述有目標，有針對性。有點兒像運動員的陪練。零假設最終的結局可能是被修正或徹底推翻。

找個栗子的話，我覺得核心假設的意思在國家隊沖金隊員不要多；這些隊員的陪練，不能菜。

我承認，由於Zuckerman的研究是在人文社科類，我在理工，有可能理解錯誤。

推薦閱讀：