学术写作
实验结果跟教材对不上?Discussion别写道歉信
Discussion不是给数据辩护,是给你自己的操作做尸检。具体性杀死'human error'式的耸肩。
凌晨一点四十七分。你盯着数据表,吸光度峰值在520nm,不是protocol里写的620nm。你的假设刚在分光光度计里凉透了。然后你准备第三次敲下'this discrepancy may be attributed to experimental error'。停手。这句话是学术界的白旗加蝴蝶结。你导师从1987年起就在每一份lab report里看到过它。他不要道歉,他要尸检报告。给他刀,别给他纸巾。
没人跟你说的那件事:Discussion不是辩护词。不是'虽然数字不对但我还是对的'。它是诊断书。导师早就知道教材答案是什么,他知道hypothesis本该成立。他打分的不是你的结果,是你的脑子能不能把一个翻车的结果拆成具体的、站得住脚的原因。一个处理得当的null result,分数比一个完美结果配两句'this supports the hypothesis'高得多。Discussion是你证明你能思考的地方,不是记录的地方。
'结果不对'的discussion,用四拍结构。第一拍:一句话说清偏差,别煽情。'The observed absorbance was 520 nm, diverging from the predicted 620 nm.' 完事。第二拍:给两三个具体的、有机制的解释。不是'errors occurred'。哪个变量偏了?温度比protocol高了2度?buffer浓度抽错了?反应时间少了40秒?说机制,别说感觉。第三拍:至少挂一个外部文献。别的组有没有报道过类似偏移?这个chromophore在文献里有没有已知的edge case?第四拍:一句'如果重做我会改什么'。四拍。不是七段焦虑式碎碎念。
看措辞。这三组改写才是真正落地的:❌ 'The results did not match the expected outcome, which may be due to experimental error.' → ✅ 'The absorbance peak at 520 nm, rather than the predicted 620 nm, suggests incomplete oxidation, likely because the mixture was held at 37°C for 12 minutes instead of the protocol's 20.' ❌ 'This discrepancy could be because of human error in measuring the volume.' → ✅ 'A 0.5 mL overpour in the aliquot transfer would shift concentration by ~4%, consistent with the 3.8% deviation observed in well B4.' ❌ 'In conclusion, the hypothesis was not supported, but this may be due to limitations.' → ✅ 'The 12-minute incubation, 40% under protocol, is the most probable cause; extending it to 20 minutes in a repeat trial would isolate whether time, not technique, drove the discrepancy.'
'human error'是Discussion里的自杀陷阱。你第三次写'experimental error'的时候,已经告诉导师你停止思考了。你把分析外包给了一个万能短语——它解释了一切,什么都没证明。具体性就是全部。'I pipetted wrong'是认罪。'The 10 µL tip was likely saturated beyond capacity at 2:30, introducing a 3 µL overshoot into well B4'是科学。前者是耸肩,后者是机制。如果你真的不知道哪里错了,就承认——但承认的方式是对你自己操作过程的假设,不是blanket excuse。
实操:先把四拍写成bullet points,写得丑没关系。然后如果draft读起来像法律免责声明——每句话都在hedging,每个claim都埋在'may''could''it is worth noting that'底下——丢进easydue跑一遍,让它标出你藏在qualifier后面、不敢commit的具体原因。你不需要显得确定,你需要显得形成了判断并且能defend。工具不会替你写机制,但会抓住你第五个'perhaps'——你本来想说的是'buffer浓度偏了'。
你的数据不需要好看。你的思考需要可见。写'my results were wrong'的学生拿B+。写'my results were wrong, and here is the specific three-step chain explaining why, and here is how I would test it next week'的学生拿A——有时候更高,因为导师刚看着你实时思考。Hypothesis从来不是重点。你的诊断能力才是。现在合上电脑,去睡。数据不会跑,ddl也不会跑。