IB MYP科学 Criterion C 数据处理 结论评价
MYP 科学的实验报告里，Criterion C 的正式名称是 Processing and evaluating，满分 8 分，与 A、B、D 三项并列。很多学生把时间花在写步骤和结论上，等报告发回来才发现丢分最多的是数据表、图表和方法评价这三处。IB《Sciences guide》要求每个 objective 的所有 strand 每学年至少各评两次，也就是说同一套技能会在不同实验里反复出现，早点写顺，后面每次都能省力。下面按五条 strand 的顺序拆开讲，每条给 Year 5 的达标要求、学生常写的版本与改写方向。
In the IB Middle Years Programme, Criterion C carries the official title Processing and evaluating and is marked out of 8, the same maximum as the other science criteria. It begins where Criterion B ends: the method is fixed and the readings are in the book, so the marks now depend on how honestly you present, process and judge them. The Sciences guide requires every strand of all four objectives to be assessed at least twice each year, which is why the same report skills keep returning. The sections below follow the five strands in order and give the wording that lifts a response out of the middle bands.
一、Criterion C 的五条 strand 与 8 分制
Criterion C 的五条 strand 依次要求：正确收集、整理、转换与呈现数据；准确解释数据并用科学推理说明结果；依据实验结果检验假设是否成立；评价方法是否有效；说明能改进这项实验的改动或延伸。MYP 科学四项标准各最高 8 分，合计 32 分，再折算成 1 至 7 的等级。描述词随年级收紧，Year 1 只要求 discuss a prediction 与 describe improvements，Year 5 升级为 evaluate the validity of a hypothesis、evaluate the validity of the method 和 explain improvements。IBO 官网公布的《Sciences guide》(for use from September 2014/January 2015，2017 年 9 月更新) 另有一条硬规定：每学年学生必须独立完成一次同时按 B 与 C 评估的科学探究。
Criterion C is one of the four sciences criteria, each with a maximum of 8, giving 32 in total before conversion to the 1-7 grade. Its five strands ask you to collect, organize, transform and present data, to interpret it with correct scientific reasoning, to evaluate whether the hypothesis held, to evaluate the method itself, and to explain improvements or extensions. The descriptors tighten as you move up the programme, so the command words that earn marks in Year 5 are stronger than those in Year 1. One further rule is easy to miss: in every year of MYP sciences, every student must complete a scientific investigation assessed against both criteria B and C.
二、原始数据与转换后的数据
strand i 里的 transform 是最常被跳过的一个词。原始读数是仪器直接显示的东西，比如三次称量得到的 2.41 克与 2.39 克；转换后的数据是你算出来的量，比如三次读数的均值、质量除以体积得到的密度、周期平方后的 T²。两类数据可以并排出现在同一张表里，但必须分成不同的列，读者要能一眼分清哪一列是记录、哪一列是计算。转换通常有三个目的：把随机误差压小、把曲线拉成直线、把量纲换成能用斜率读出物理量的形式。很多学生以为画出原始数据就算完成了数据处理，IB 要看的恰恰是你有没有主动做这一步。
Raw readings are what the instrument showed; transformed data are the quantities you calculate from them, such as the mean of repeats, a density obtained by dividing mass by volume, or the square of a period; keep the two kinds in separate columns so a reader can tell recorded values from calculated ones at a glance. Transformation is usually done for one of three reasons: to reduce the effect of random scatter, to turn a curve into a straight line, or to arrange the quantity so its gradient carries physical meaning. A table of untouched readings, with nothing calculated from it, leaves strand i half finished.
三、结果表格的规范
结果表至少要交代三块信息：自变量的一列、每一组重复读数、以及由重复算出的均值与转换量。单位写在表头而不是每个数字后面，写成 Length / cm 这种形式；同一列的有效数字位数保持一致。列的顺序通常把自变量放在最左边，读表的人习惯从左往右看趋势。常见的错误是把两类变量挤进一列，或者把均值写到仪器给不出的位数：2.41、2.39、2.44 三个读数的均值是 2.41，写成 2.4133 就把天平的精度夸大了。表格还需要一句能独立读懂的标题，例如 Table 1: period at five lengths (mean of three trials)。
A results table needs three blocks of information: the independent variable, every repeat reading, and the mean or transformed value calculated from those repeats. Units belong in the column heading, written as Length / cm, rather than after every number, and all values in one column should carry the same number of significant figures. Averaging 2.41, 2.39 and 2.44 g gives 2.41 g; reporting 2.4133 g claims a precision the balance never had. A caption such as Table 1: period at five lengths (mean of three trials) tells the reader what the numbers represent before they start on the trend.
四、图表：自变量放横轴，用最贴合的线
图表的判断标准比表格更硬：自变量放横轴、因变量放纵轴，两条轴都写清物理量和单位，刻度均匀，数据点铺满图面的大部分而不是挤在左下角。散点画完之后用一条直线或平滑曲线去拟合，不要用直线把相邻的点连起来，点对点连线会把噪音画成趋势，也让你无法从斜率读出任何东西。如果理论预测成正比，就把纵轴换成应该线性的那个量，例如用 T² 对 L 作图，再看拟合线是否穿过原点。拟合线明显不穿过原点，本身就是一条值得写进评价的信息，不需要抹掉。异常点换一种符号标出来，并在正文里解释它的来源。
Graph marks follow conventions that are quick to check: the independent variable on the horizontal axis, the dependent variable on the vertical one, both axes labelled with quantity and unit, and scales chosen so the points fill most of the grid. Plot the points, then fit a single straight line or smooth curve instead of joining point to point, because a chain of segments turns random scatter into a fake trend and leaves nothing to read from a gradient. When theory predicts proportionality, plot the quantity that should be linear, such as T squared against L, and check whether the line passes through the origin. A clear intercept is evidence worth discussing rather than a flaw to hide, so mark any anomalous point with a different symbol and explain it in the text.
五、用科学原理解释趋势
strand ii 要的是解释，不是描述。写出长度增加时周期变长，只到 3-4 档；要到 5-6 以上，必须把趋势与物理关系连起来。以单摆为例，周期 T 与摆长的平方根成正比，T = 2π√(L/g)，摆长变成四倍，周期只翻一倍；画 T² 对 L，斜率就是 4π²/g，可以用它反算 g 再和当地重力加速度比较。化学实验里对应的说法是浓度加倍使初始速率约翻倍，说明该反应对这一反应物是一级。解释里出现的每个符号都要带单位，公式里的每个量都要和表格中的自变量、因变量对得上，否则评语会写 reasoning not linked to the data。
Strand ii asks for explanation rather than description, so a sentence such as the period grew as the string got longer stays in the 3-4 band. To move up, connect the pattern to the relationship theory predicts: for a simple pendulum the period varies with the square root of the length, so quadrupling the length doubles the period, and plotting T squared against L gives a gradient of four pi squared over g from which g can be recovered and compared with the accepted value; in chemistry the equivalent move is to argue that doubling the concentration and doubling the initial rate indicates a first-order dependence on that reactant. Every symbol should carry a unit and match the variables in your table.
六、检验假设：支持、部分支持与推翻
检验假设要先分清三种结果：被数据支持、被部分支持、被推翻。直接写假设正确，只到 3-4 档；7-8 档会说清哪一部分成立、哪一部分不成立，以及依据在哪。假设摆长增加时周期按平方根关系增加，如果 T² 对 L 的图是过原点的直线，方向与形式都成立，假设整体被支持，你可以引用斜率与理论值的接近程度作为依据。如果斜率与理论值差出 15% 以上，比较诚实的写法是说明方向成立而数值关系存疑，再把这个差距交给 strand iv 解释，而不是回头改数据。切忌在结论里写下数据支撑不了的话，阅卷时这类句子会被单独圈出来。
Judging a hypothesis comes down to three honest outcomes: supported, partly supported, or rejected. Saying the hypothesis was correct sits in the 3-4 band, while a top-band answer states which part of the prediction held, which part did not, and what evidence decides it. If the graph of T squared against L is straight through the origin, both the direction and the form of the prediction hold, and the closeness of the gradient to the theoretical value is the evidence you cite. Where the gradient differs from theory by more than about 15 per cent, the defensible claim is that the trend was right but the numerical relationship is uncertain, and that gap belongs in the evaluation of the method rather than in a rewritten set of readings.
七、方法评价：随机误差与系统误差
评价方法的关键在于具体。只写存在人为误差，说明不了哪一步、哪个读数、影响朝哪个方向，因而上不到 5-6 档。好的评价会分开两类误差：随机误差让读数在真值两侧散开，重复测量取均值可以压小它，常见来源是反应时间和读数视线；系统误差让所有读数朝同一方向偏，取均值没有用，常见来源是零点未校准、尺子刻度磨损、温度计浸入深度不足。误差还要带上大小：用秒表计时十次摆动，可以把反应时间的影响压到十分之一；目测判断摆球中心，会让摆长带上约 1 毫米的不确定度。最后把它落到你的数据上，比如 0.60 米那个点偏高，很可能就是那次放开摆球的角度过大。
Specificity separates a substantial evaluation from a vague one, and the phrase human error on its own says nothing about which step, which reading or which direction. Distinguish random from systematic effects: random scatter falls when you average repeats, while a zero error on a balance or a worn scale shifts every reading the same way and survives averaging. Attach a size to each weakness, for instance timing ten swings reduces the reaction-time contribution roughly tenfold, and measuring to the centre of the bob by eye adds about a millimetre to the length; then link the weakness to a particular result, such as the point at 0.60 m sitting above the line because the bob was released at a wider angle on that run.
八、改进与延伸：一条改进对应一条缺点
strand v 的评分点在于改进与前面指出的缺点是否一一对应。前面说反应时间是主要误差，后面就写用光电门或电子计时器替代秒表；前面说摆长靠目测，后面就写用游标卡尺量到球心；前面说释放角度不稳定，后面就写用标好角度的卡片固定起始位置。每条改进还要写清它改变了什么，例如把计时的离散度从 0.2 秒降到 0.01 秒。延伸是另一个方向的问题：换成不同质量的摆球，检验周期与质量无关的预言；或者把同一套方法用到弹簧振子上，比较两种周期公式。写一堆与缺点无关的改进，或者只写下次更小心，都只停在 describe 的档位。
Improvements earn marks when each one answers a weakness you have already identified. If the reaction time dominates the uncertainty, a photogate and electronic timer is the matching improvement; if the length is judged by eye, a vernier caliper taken to the centre of the bob answers it; if the release angle varies, a card marked with the starting angle fixes it. Say what each change does to the uncertainty, for example cutting the timing spread from about 0.2 s to about 0.01 s; extensions move in another direction, such as repeating the investigation with bobs of different mass to test the prediction that mass does not affect the period; a long list of unrelated suggestions stays at the describe level.
九、异常点：解释而不是删掉
异常点是偏离拟合线的那一个点，处理它有三种做法，只有一种能拿分。直接删掉，评委会认为你在挑选数据；留着不解释，说明你没有回看自己的数据；把它标出来并给出可检验的原因，才是 7-8 档的做法。可接受的原因通常落在方法上：那一次释放角度太大，超出了小角近似成立的范围；那一次称量时容器外壁带着水；那一次读数时温度还没稳定。除了给出原因，还要说明它对结论的影响有多大：保留或去掉这个点，斜率的变化都在不确定度范围之内，结论不变。这句话写出来，说明你分得清单个数据点与整体结论的关系。
An anomalous point is the one sitting away from the fitted line, and there are three ways to handle it. Deleting it looks like selecting data to suit the conclusion, while leaving it unexplained suggests you never looked at your own results; marking it and offering a testable reason is the top-band move. Acceptable reasons usually point back at the method, such as a release angle large enough to break the small-angle approximation, water on the outside of a weighing container, or a temperature that had not settled. Then say what difference the point makes: if the gradient moves by less than the uncertainty whether the point is kept or removed, the conclusion stands.
十、不确定度、有效数字与误差条
到了 7-8 档，评委会期待你对不确定度有基本处理。最简单的做法是给每个测量量写一个正负值：米尺正负 1 毫米，秒表正负 0.2 秒，量筒正负 0.5 毫升，来源是量具的最小分度或它的一半。算出来的量要做传递：密度等于质量除以体积，质量的相对不确定度是 0.4%，体积是 0.8%，密度的相对不确定度大约是两者相加的 1.2%。有效数字要与这个不确定度一致，均值的不确定度是 0.01 克时，2.41 克写成 2.410 就是过度声称精度。误差条按同样的数值画在图上，误差条穿过拟合线的点，不算异常点。
At the top band, examiners expect some treatment of uncertainty rather than bare numbers. A workable start is to write an uncertainty beside each measured quantity, taken from the smallest division of the instrument or half of it: 1 mm for a metre rule, 0.2 s for a stopwatch, 0.5 ml for a measuring cylinder. Uncertainties combine when quantities are calculated, so a density from a mass known to 0.4 per cent and a volume known to 0.8 per cent carries roughly 1.2 per cent. Significant figures should match that uncertainty, and error bars drawn with the same values show whether a point really sits off the line.
十一、动词决定分档：state 到 evaluate
strand 描述里的动词就是评分路标。state 只要说出结论，outline 要给一句理由，discuss 要摆出正反两面的证据，evaluate 要下判断并说明依据。Year 1 的描述词用 discuss the validity of a prediction，Year 3 换成 discuss the validity of a hypothesis，Year 5 进一步要求 evaluate the validity of a hypothesis 与 evaluate the validity of the method。同一份报告在 Year 3 可能拿到 6 分，到了 Year 5 只剩 4 分，差别通常不在数据，而在用词的深度。一个实用的自检办法：逐句读自己的评价段，看每一句能不能回答依据是什么，答不上来的句子只是描述，不是评价。
The command words in the descriptors work as signposts for the marker. State means give the conclusion, outline means add one supporting reason, discuss means weigh evidence in both directions, and evaluate means reach a judgement and justify it. Year 1 asks you to discuss the validity of a prediction, Year 3 to discuss the validity of a hypothesis, and Year 5 to evaluate both the hypothesis and the method. A report that scores well in Year 3 can fall two levels in Year 5 without any change in the data, because what is compared is the depth of the judgement. A quick self-check is to read each evaluation sentence and ask what evidence supports it.
十二、报告里反复出现的失分点
我在批改 MYP 科学报告时，错得最多的几处相当固定：用点对点连线画图，趋势被锯齿线盖住；单位重复写在每个数字后面，或者均值的小数位比仪器能给的多；结论只复述数据，没有回到科学原理；评价里写人为误差却不说哪一步；改进与缺点对不上，例如读数误差却建议换一支更贵的秒表；异常点悄悄消失；结论里的因果超出实验能支持的范围，两个变量同步变化就直接推出因果关系。这七处每次都能见到五六处，拿自己的稿子逐条对照删改，比重新写一遍省时间。
A handful of faults turn up in almost every batch of MYP science reports. Graphs joined point to point hide the trend behind a sawtooth line; units are repeated after every number instead of sitting in the heading; means carry more decimal places than the instrument can justify; conclusions restate the numbers without returning to the science; evaluations name human error without saying which reading it spoiled; improvements do not match the weaknesses listed; anomalous points quietly disappear; and a conclusion claims cause and effect from two variables that merely moved together. Checking your own draft against that list usually takes less time than rewriting it.
十三、交稿前的十分钟冷读
交稿前留十分钟做一遍冷读。先只看表格与图，问自己三件事：单位在表头吗，每个数据点都有对应读数吗，拟合线是用尺子或软件画出来的吗。再读结论段，确认每一句都能在数据里找到依据，并且回到公式或原理。最后读评价段，用笔划出每一条改进，看它是否对应前面的一条缺点；对不上的删掉，宁可只留两条真正相关的改进。整篇报告的页码、图表编号与假设的表述保持一致，假设那句话最好原样抄进结论与评价，评委不必回头找。改完放一放再读一遍，你能看出前一天没看出来的问题。
Leave ten minutes for a cold read before submitting. Look at the tables and graphs on their own and check three things: units in the headings, a plotted point for every reading, and a fitted line drawn with a ruler or fitted by software. Read the conclusion next and confirm that every claim traces back to a number in the data and to a principle or equation in the syllabus. Then go through the evaluation and underline each improvement, keeping only those that answer a weakness you actually named, and make sure the wording of the hypothesis is identical wherever it appears, so a moderator never has to guess which version you tested.
TI-Nspire CX II CAS 图形计算器 AP/IB/A-Level¥1320📱 去京东购买京东精选
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
16621398022 同微信
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图