Jie Shen
Experimental archaeology is … an important way in which archaeology can come to grips with the mental processes, plans and activities of prehistoric man.
-Coles 1966
I’ve been doing different replication experiments since my master’s: first basket impressions on pottery, then bone tools, stone tools and all sorts of animal products. Especially in the past two years, after gaining some experience through a pile of useless experiments, I’ve finally started producing data with archaeological meaning and practical value. I’m not a skilled craftsperson yet, but all this tinkering has made me something of a locally famous artisan among my friends. After useful, useless, successful and (many) failed experiments, I’d like to share some personal experience. It’s very subjective and mostly consists of pitfalls I’ve fallen into. I hope friends thinking of stepping into this interesting and slippery pit called experimental archaeology can read this, fall over a few fewer times, waste a little less material, and eventually produce plenty of results.
I. The Most Important and Most Easily Overlooked Point: Doing It Is Harder Than Imagining It.
Experiments always go beyond what the design anticipates. A design is a theoretically feasible plan. Even if we’ve exhausted the historical sources, ethnographic records and other research we can find, some details won’t have been recorded. Take preparing bone tool blanks: find a long bone, cut off the epiphyses, split the shaft. Such a simple plan can run into:
“The bone I bought was bleached, losing too much organic matter and becoming too brittle to work.”
“The stone knife got halfway through, but it’s too thick to cut any farther into the bone; I need to thin the knife first.”
“I’ve finally got just the shaft left, but it isn’t an ideal four-sided prism: it’s a twisted four-sided prism plus a three-sided prism. So how do I split it with the least material wasted?”
In reality, we face far more variables than expected. Materials, tools and participants may all be less than perfect, and sometimes results exceed expectations too. Plans matter, but change outpaces plans. Be ready to adapt to reality and revise the approach.
Actual work is never as simple or smooth as imagined either. Again with bone blanks, I gave my students flint knives to try cutting and wedge splitting. Eight people worked for an hour and a half, averaging only one cut each. It wasn’t lack of strength: they weren’t familiar with stone and bone properties, weren’t used to handling the tools, and didn’t know the most comfortable grip or way to apply force. Even when the method and design are sound, whether we can carry them out well matters. If a method that has worked for others or should have a high chance of success doesn’t work, first consider your own execution or proficiency. Often, sticking with it a few more times turns things around. Then consider mistakes in the process, and only finally consider switching methods.
People always make mistakes. We buy the wrong materials, misremember numbers, forget to time things, lose records and photos, forget what we read, get stuck on simple practical problems, get tired or bored, and skip the planned procedure to save trouble. I’ve made all these mistakes and suffered for them (and, stubborn as ever with my terrible memory, keep making them, haha). Mistakes can come from you, experimental partners, or NPCs who have nothing to do with the experiment. So expect mistakes, accept them, deal with them, and keep an eye on each other. Don’t wreck the experiment together.
II. Always Do Pilot Experiments.
The way to deal with all those chaotic evil forces above is pilot experiments. This is very important: you test whether the procedure works, develop your own proficiency, and gather data along the way, getting a better handle on effectiveness, efficiency and expected time. We all know novices and experienced people can affect results very differently. Becoming proficient before the formal experiment reduces errors caused by skill differences. In bone-working experiments, for example, I’m much more efficient than my students, not only because I know what’s right, but because I know which operations are wrong and what consequences they might have. Between those two, my efficiency is much higher.
III. Leave Some Slack in Your Plan.
When buying materials, it’s best to prepare twice as much as expected. This comes from recent experience too: you never know which step will use more than anticipated, and sometimes something you thought unnecessary turns out to be in high demand. So get a little extra. Leftovers can be used next time; shortages really do waste time.
When estimating experiment duration, add at least a third as a buffer. That’s for mistakes, surprises and extra experiments. If everything really goes smoothly, you can write the report while it’s fresh.
IV. Know the Properties of Different Materials.
Some studies’ interpretations of artefact functions are unrealistic or insufficiently rigorous fundamentally because they don’t understand the materials’ properties and make assumptions about their uses. As someone born and raised in a city, I originally knew little about traditional craft materials or tools. Only after doing experiments for a while and encountering more materials did I gradually feel I’d built up something like “common sense”: a rough understanding of what materials can be worked by which methods and to what effect. This “understanding” can’t come from books or the internet, only from actually interacting with things in the real world.
V. Be Clear About Your Questions, Hypotheses and Controlled Variables
Set your research question first: are you testing a relationship between a process and an outcome, comparing outcomes from different processes, or exploring possible causes of an outcome? The same action may produce different results because of uncontrollable factors, and several different actions may produce the same result. So identify the key factors you want to verify or test when setting the question.
Experiments often contain far more variables than we imagine: raw material types and condition, preparation methods and duration, tools, working methods, the worker’s proficiency and ability, use methods and duration, and so on and on. To obtain relatively stable results or establish each variable’s effects, control variables as much as possible, ideally changing only one per experiment and comparing results afterwards. More variables may mean more experiments.
But don’t start by designing a huge experiment covering many variables: it costs money and time, and results probably won’t be very clear. Such experiments can be achieved gradually in stages, but aren’t suited to doing everything at once as a student or when short of people. I once made an ambitious plan to test working results with hundreds of bones under different processes and materials. The venture died before it even began: if every variable combination needed at least five repeats for stable results, I simply couldn’t afford that many bones, or find the time and energy.
Before and during the experiment, keep asking what you ultimately want to demonstrate. Do you want a successful result, or to test a method? If you’re testing a method, follow the original design from beginning to end. Even if the craft technique doesn’t produce the desired effect, that failure or shortfall is valuable. If you want a successful result—making an object or achieving an effect—don’t rigidly stick to the design; improve the method as the experiment and the material’s condition develop. But carefully record the variables you change for future reference.
During experiments, I often encounter disappointing situations: tools aren’t as handy as expected, or materials don’t change as imagined. I get anxious or start doubting my original idea and immediately want another route. That’s when I most need to question myself: is my priority testing whether this technique works, or finishing the object? Even if finishing is the priority, can I keep going a little longer to see whether the method is inadequate, or whether I’ve only just started and the expected effects haven’t appeared yet? Sometimes I find the method does work if I persist. I just lacked experience, wasn’t doing it well enough, or it simply needed time to show results.
If you start chasing “completion” and only want a result, regardless of whether variables are controlled or the process stays close to the original context of use—whether by overusing modern tools or speeding up what should be steady work out of impatience—being blinded by the “result” leaves you with something resembling the ancient example, but a process that can’t explain anything.
VI. Having an Experienced Person Guide You Really Matters.
Sometimes the problem with replication isn’t knowing what to do, but not knowing how to fix a problem, or whether there’s a problem at all. Take hide tanning, a process combining physical and chemical changes. I consulted lots of videos, books and ethnographic records, nearly all of which only documented successful procedures. Countless times during my experiment, I’d finish a step and notice the result didn’t seem like the book or video—but I didn’t know what I’d done wrong or not done enough of. By the end, with errors piling up, the tanning had pretty much failed. After that painful lesson, I signed up for a class. Only then did I realise how much know-how the sources hadn’t taught me. With an instructor explaining and analysing every step, my earlier puzzles turned out to have so many solutions. Like any craft, replication experiments benefit enormously from an experienced craftsperson’s guidance: it saves huge amounts of trial and error and hitting walls.
VII. Keep Records! Keep Records! Keep Records!
Keep records throughout: writing, voice logs, photographs and videos are all useful. Especially video: a wearable action camera records the process as well as spoken observations you don’t have time to write down. Better too much than too little, too detailed than too brief. Useful or not, repetitive or not, recording it is better than not.
VIII. Write the Report Quickly!
It’s best to write a report as soon as the experiment ends. This really matters: many ideas and insights arise during the experiment, and without prompt recording they may fade with time along with your feel for the work. If the procedure changes, it’s also easy to forget those details when reviewing it later. So however tired you are or inconvenient it feels, organise the results straight away. It needn’t be a polished article; even a simple procedure log and reflection will be very useful.
IX. Enjoy the Process
Experiments are full of strange stories and ridiculous accidents. They can be tiring, painful, confusing and sometimes very smelly. But they’re also great fun. They help us reconsider the material world surrounding us and shaping our lives from another angle, and reawaken our sense of gaining knowledge and experience beyond screens. The friendships formed through experiments can be as strong as those between people who’ve battled through the same trench. So I hope everyone enjoys the process, whatever the results.
If you’d like to start doing replication experiments, these resources may help:
Reading List | A Review of Experimental Archaeology Theory and Methods
Experimental Archaeology | An Incomplete Guide to Video and Written Resources 1.0
实验考古 | 复原实验设计和实操非正式指南
沈劼 · 2026/8/23 10:10
中文原文
Experimental archaeology is … an important way in which archaeology can come to grips with the mental processes, plans and activities of prehistoric man.
-Coles 1966
从硕士开始，我就在做不同的复原实验。最开始是陶器篮纹，然后是骨器石器，各种动物制品。尤其是最近两年，在依靠一堆没用实验积累了一些经验之后，终于开始产出一些有考古学意义的，有应用价值的实验数据。虽然还不算熟练工匠，但也靠捣腾的一堆东西在我的朋友里混成了远近闻名的手艺人。经历过了有用的、没用的、成功的、（很多）失败的实验之后，我打算来分享一些个人经验。这些经验非常的主观，基本都是我踩过的坑。希望有志于踏入实验考古这个又有趣又光滑的大坑的朋友们，看到我的经验，可以少摔几跤，少浪费一点材料，最终出很多成果。
一、最重要也是最容易被忽略的一点：做得不如想的好。
实验过程永远会超出实验设计的构想。因为设计是一个理论上可行的方案，即使穷尽了我们能查到的历史文献、民族学资料和其他研究，也还是会有没有被记载的细节。以骨器开料为例，找一根长肢骨，切掉骨骺，把骨干劈开成。如此简单的实验计划，会遇到
“买来的骨头被漂白过，丢失了过多的有机物，因此变得过脆不适合加工”，
“石刀切了一半因为太厚而无法继续切进骨骼，要先把石刀磨薄”，
“好不容易只剩骨干了，但它并不是一个理想的四棱柱，而是一个扭曲的四棱柱加三棱柱，那到底怎么劈开才是最不浪费材料的方案？”
因此，在现实中我们会面对比预想多得多的变量，材料、工具、执行者可能都并不如预期的完美，结果有时候也会超出预期。所以计划很重要，但计划赶不上变化，最好是做好随时适应现实，修改方案的打算。
实际操作也永远不会像想象的那么简单和顺利。还是以骨器开料为例，我给我的学生们发了燧石石刀，让大家尝试切割和楔裂法。八个人用了一个半小时，平均每人只切开了一次。倒不是力气不够，而是不熟悉石和骨的物性，不适应使用工具的方式，不知道怎么抓握和发力最舒适。当方法和设计都没有问题的时候，执行力是否足够，也是影响实验考古结果的一个重要因素。当某个别人成功过或者理论上成果概率很大的方法不成功的时候，先思考一下是不是自己的执行力或者熟练度问题，往往多坚持几次就会出现转机。然后再考虑是不是执行过程中有错，最后才考虑更换别的方法。
人永远都会犯错。会买错材料，会记错编号，会忘记计时，会丢失实验记录和照片，会忘记文献里看到的内容，在实际操作的时候被很简单的问题困住，还会累，会厌倦，会为了省事而不遵守既定的实验流程。这些都是我犯过的错，我也都吃了应该的苦（并且死性不改忘性过大地继续犯错hh）。错误可能不只是发生在自己身上，也可能在实验伙伴身上，在和实验无关的NPC身上。所以准备好会有错误，接受错误，应对错误，互相监督，不要一起毁掉实验。
二、一定要做先导实验。
应对前面提到的诸多混沌邪恶力量干扰的方式，就是做先导实验。这一点非常重要，不仅是先测试实验流程是否可行，也是培养实验者自己的熟练度，还能顺便积累数据，对实验效果、效率和预期时间会有更好的把握。毕竟我们都知道，新手跟老手对实验结果的影响很大，在正式实验之前先把自己变成熟练工，能够有效地减少熟练度差异带来的数据误差。比如同样是骨器加工实验，我加工的效率比我的学生们高得多，不仅是因为我知道怎么做是对的，还因为我知道哪些操作是错的，可能带来什么样的后果。一进一出，我的加工效率会高出一大截。
三、制定计划时要有余量。
买实验材料的时候，最好按预期的两倍准备用量。这也是近期实验的经验之谈，你永远想不到哪一步会比你预想的更费材料，偶尔也会发现本来以为没用的材料实际上需求很大，所以多备一点吧，用不完可以下次用，少了就真的耽误时间了。
预估实验时长的时候，最好加上三分之一及以上的余量。余量是留给失误、意外、以及追加实验的。如果真的一切顺利，正好能趁热打铁写实验报告。
四、要了解不同材料的性质。
有的研究对器物功能的推测缺乏实际或者不够严谨的根本原因，其实是不够了解实验材料的物性，所以对其应用方式有一些想当然的推测。我作为城市出生生活的小孩，其实本来业对传统手工业的材料和工具都没有什么认识。但实验做久了，接触过更多的材料，才慢慢感觉自己积累起来一些“常识”一样的经验和知识，并且能够大致了解什么材料可以通过什么方法加工，能够达到什么效果。这种“了解”，并不能通过书本和网络得到，只能从自己真实和现实世界中的物质交互过程中积累。
五、明确实验问题、假设和控制变量。
实验前先确定好研究问题，是要验证某个过程与结果的关系，还是比较不同过程的结果，或者是探索某一结果可能的成因。很多时候同一个行为可能会因为含有不可控因素而产生不同的结果，或者好几种不同行为可能会导致同一结果，所以在确定研究问题时，就要想好自己要验证或者测试的关键因素是什么。
实验中所包含的变量往往比我们想象的多很多，原材料的种类和状态、材料处理方式、处理时长、加工工具、加工方法、加工者的熟练度和能力、使用方式、使用时长等等等等。要得到一个相对稳定的结果，或者要探索清楚每一个变量对结果的影响，就要尽量控制变量，尽量每一次实验仅改变一个变量，最后再来对比结果。所以涉及的变量越多，实验次数可能就越多。
但是，不要一开始就设计一个规模宏大覆盖不同变量的实验，因为费钱、费时间、结果大概率不会很显著，这样的实验适合分阶段逐步实现，但不适合学生阶段或者实验人手不足的阶段一口吃个胖子。我曾经做了个雄心壮志的计划，要用几百块骨头测不同过程和材料的加工结果，创业未开始就中道崩殂，因为我发现其中每一个变量组合都需要重复至少五次来获得稳定结果的话，我实在是买不起那么多骨头，也没有这个时间和精力来完成。
实验前和实验过程中都要想清楚一个问题，就是最终要证明什么。是想要结果成功，还是要测试方法。如果是要测试方法，那就是从头到尾都要坚持最初的实验设计，哪怕要测试的工艺或者加工方式无法达到最终效果，这个失败或者不足的过程也是有价值的。如果是要一个成功的结果，比如做出某个物品、达成某种效果，那就不能一直死守实验设计，要随实验过程和实验对象的状态改进方法。但与此同时也要好好记录改变的变量，方便以后参考。
我在做实验的过程中时常会遇到不如人意的情况，可能工具不如预想的趁手、材料没有如预想的改变，这种时候往往开始着急或者否定自己原先的设想，立刻就想换一条路走。但这种时候最需要拷问自己：我到底要优先测试这个技术是否有效，还是要优先完成这个东西的制作？就算是要优先完成制作，是不是可以再坚持一下，看到底是现有的方法不足，还是因为我刚开始做，所以应有的效果还没有显现出来？偶尔我会发现继续坚持做下去，现有的方法其实也是有效的。只是因为最开始我的经验不足，执行得不够好，或者就算需要一段时间才能出效果。
如果做着做着开始追求“完成”，只想要某个结果出现，而不管过程中是否有很好的控制变量，或者是否尽量贴近原有使用场景。不管是过多使用现代工具，还是本来匀速加工因为缺乏耐心变成高速加工，被“结果”蒙蔽了心智的结果，就是看起来做出了和古代样本差不多的东西，但是过程并不能用来说明任何问题。
六、有熟练工带着真的很重要。
复原实验有时候不是不知道怎么做，而是出问题时不知道怎么解决问题，或者不知道有没有出问题。以鞣皮实验为例，鞣皮是个综合了物理和化学变化的加工过程，我查了大量的视频、书和民族学记录，它们几乎都只记录成功的过程。实验过程中我无数次意识到，这一步做完，我看到的结果好像和书里或者视频里描述的不一样，但是我并不知道是哪儿做得不对，或者做得不够。等到这样错上加错的做完的时候，鞣皮已经几乎失败了。后来我痛定思痛，报了个班去跟着老师学的时候，我才意识到，原来过程中间有如此多的know-how我未曾从资料中学到，而有了老师之后，每一步都有解释和分析，我之前所困惑的问题竟有如此多的解法。复原实验就像每一门手工业技术一样，老师傅的点拨真的能省去海量自己摸索撞墙的时间。
七、留记录！留记录！留记录！
实验过程保持记录，文字记录、语音日志、拍照、摄像都很有用。特别是摄像，一个随身的运动相机既能记录过程，又能记下实验中来不及写的口述内容。记录宁多勿少，宁繁勿简，不管有没有用，重不重复，记下来总比没记好。
八、快写报告！
最好能实验结束赶紧写一个报告出来。这一点真的很重要，因为很多思考和经验是在实验过程中产生的，如果没有及时记录下来，它们可能就和你的手感一起，随着时间流逝而消失了。而且有时候实验过程如果有变动，之后复盘再梳理实验流程的时候，很容易会忘记更改的细节。所以实验结束后，不管再累再麻烦，最好也能立刻整理一下实验结果，不用写一个很完善的文章出来，哪怕是一个简单的流程记录和反思，也会非常有用。
九、享受过程
实验过程会充满了离奇的故事、啼笑皆非的事故，会疲惫、痛苦、迷茫，有时候还会很臭。但实验也会非常有趣，会帮我们从另一个角度重新审视这包围我们和塑造我们生活的物质世界，重新唤起我们对屏幕以外知识与经验积累过程的感知。共同做过实验的人之间的情谊，往往不亚于在同一个探方里奋战的方友。所以希望大家，不管结果好坏，都能享受实验过程。
如果你想开始复原实验之旅，下面这些资料可能会有帮助：