📚 Identifying Limitations in Procedures and Suggesting Improvements | 实验步骤局限性识别与改进建议
In CIE A-Level Physics practical papers, especially Paper 3 and Paper 5, evaluation marks are awarded for identifying genuine limitations in an experimental procedure and suggesting improvements that are directly linked to those limitations. A good limitation is specific, observable and procedural, such as 'the stopwatch was started and stopped by hand, so every measured time included a reaction-time delay of about 0.2 s'. Vague statements like 'the equipment was not accurate' earn no credit. This article builds a complete, exam-ready framework: how the marking works, how to classify errors, a catalogue of limitations by apparatus, the exact language of a full-mark improvement, and two worked model answers.
在 CIE A-Level 物理实验卷（尤其是 Paper 3 和 Paper 5）中，评分要求学生找出实验步骤的真实局限性，并提出与这些局限性直接相关的改进。好的局限性必须具体、可观察且与步骤有关，例如 '秒表由手动启动和停止，因此每次测得的时间都包含约 0.2 s 的反应延迟'。像 '设备不够准确' 这样笼统的说法不能得分。本文将搭建一套完整的应试框架：评分机制如何运作、误差如何分类、按仪器归类的局限性清单、满分改进的确切语言，以及两篇范例答案。
1. Why Identifying Limitations Matters | 为什么识别局限性很重要
Evaluation is the part of the practical paper where the largest number of candidates lose marks, not because they cannot think of an error, but because they write a general complaint instead of a procedural limitation. The examiner needs to see a cause, a mechanism, and an effect on the measured quantity. 'The ruler was inaccurate' fails all three tests; 'the metre rule was held vertically beside the wire while the eye was level with the bench, so parallax caused the extension to be read up to 2 mm too large' passes all three.
评价部分是实验卷中失分最多的部分，原因往往不是学生想不出误差，而是写了一句笼统的抱怨，而不是针对步骤的局限性。考官希望看到三样东西：原因、作用机理，以及对被测物理量的影响。'尺子不准' 三项都不满足；'米尺竖直放在金属丝旁，而视线与桌面齐平，因此视差使伸长量读数偏大最多 2 mm' 三项全部满足。
Every improvement you propose must also be realistic in a school laboratory. It should do at least one of the following three jobs: reduce a random error, remove a systematic error, or extend the range of data so that a graph covers a wider interval. An improvement that is expensive, unavailable, or does not actually attack the stated limitation will not score.
你提出的每一条改进还必须是在学校实验室里可行的。它至少应完成下面三项任务之一：减小随机误差、消除系统误差，或扩大数据范围使图像覆盖更宽的区间。昂贵、不具备条件、或者根本没有针对所写局限性的改进都不得分。
Percentage uncertainty = (Δx ÷ x) × 100%
A useful rule of thumb is that the percentage uncertainty of the measured quantity decides how many significant figures are justified, and therefore decides which limitation actually matters. If the timing uncertainty is 0.2 s on a 20 s measurement, that is only 1%, and blaming the stopwatch without quantification is weak; if it is 0.2 s on a 0.5 s fall, that is 40%, and the same stopwatch becomes a decisive limitation.
一条实用的经验法则是：被测物理量的百分比不确定度决定了结果该保留几位有效数字，也因此决定了哪一条局限性才是真正关键的。如果 20 s 的测量中计时不确定度为 0.2 s，那只有 1%，不做量化就指责秒表是站不住脚的；如果 0.5 s 的下落过程里同样是 0.2 s，那就是 40%，同一个秒表就成了决定性局限。
2. What Examiners Actually Reward | 考官真正给分的点
Mark schemes in Paper 3 typically allow two evaluation marks per experiment, and Paper 5 awards a small block of marks for limitations and improvements in the evaluation question. The marks are almost always given for a matched pair: a limitation plus an improvement that removes or reduces it. Writing two limitations and no improvement, or two improvements with no stated limitation, usually caps the score.
Paper 3 的评分标准通常每道实验题给两个评价分，Paper 5 则在评价题中给一个小的分值组。这些分数几乎总是给成对出现的答案：一条局限性加上一条能消除或减小它的改进。只写两条局限性不写改进，或只写两条改进不写局限性，通常会被扣分封顶。
- Specific and procedural: it names the apparatus and the action. | 具体且与步骤相关：点明仪器和操作动作。
- Observable: a reader could watch the experiment and see the problem. | 可观察：读者看实验过程就能看到这个问题。
- Quantified where possible: give an estimated size of the effect. | 尽量量化：给出该影响的大致量级。
- Not already controlled in the described method. | 不是题目描述的方法中已经控制过的因素。
- Improvement is technically feasible and proportionate. | 改进在技术上可行且与问题相称。
The last point catches many candidates. If the question states that the student used a fiducial mark and timed 20 oscillations, then 'use a fiducial mark and time more oscillations' scores nothing, because the method already does it. Always read the procedure paragraph twice before writing any evaluation point.
最后一点让很多考生失分。如果题目已经写明学生使用了标记线并测量 20 个周期，那么 '使用标记线并多测几个周期' 一分不得，因为方法里已经做到了。写任何评价点之前，务必把步骤段落读两遍。
3. Random, Systematic and Zero Errors | 随机误差、系统误差与零点误差
Classification matters because the correct improvement depends on the type of error. A random error scatters readings about the true value and is reduced by repeating and averaging, or by improving resolution. A systematic error shifts every reading in the same direction and is only removed by calibration, a change of technique, or a change of apparatus. A zero error is a special systematic error present when the instrument reads a non-zero value for a true zero input.
分类很重要，因为正确的改进取决于误差类型。随机误差使读数围绕真值散布，可通过重复取平均或提高分辨率来减小。系统误差使每个读数朝同一方向偏移，只能通过校准、改变操作方法或更换仪器来消除。零点误差是一种特殊的系统误差，指输入真值为零时仪器读数不为零。
| Limitation | Error type | Correct improvement |
| Hand timing of a short interval | Random | Light gates and data logger, or time many oscillations |
| Micrometer reads 0.03 mm when closed | Zero error | Subtract the zero reading from every value, or recalibrate |
| Thermometer stem not in the liquid | Systematic | Immerse to the correct depth and stir before reading |
| Heat loss to surroundings in a specific heat experiment | Systematic | Insulate the block, or use the initial-gradient method from a cooling graph |
| Ammeter has non-zero resistance | Systematic | Correct for the ammeter resistance, or use a smaller-resistance meter |
Note that 'repeat the measurement' is only a valid improvement for random error. Repeating an experiment affected by a zero error simply produces a precise wrong answer, which is exactly what the examiner does not want to read.
注意，'重复测量' 只对随机误差是有效改进。对存在零点误差的实验重复测量，只会得到一个精确的错误答案，而这正是考官不想读到的内容。
4. A Catalogue of Limitations by Apparatus | 按仪器归类的局限性清单
Most Paper 3 and Paper 5 evaluation questions can be answered from a compact mental catalogue. Learn the limitation and the improvement together as a pair, and adapt the numbers to the experiment in front of you.
大多数 Paper 3 和 Paper 5 的评价题都可以用一份精简的清单来作答。把局限性和改进成对记住，再把数字根据眼前的实验调整即可。
- Stopwatch: reaction time about 0.2 s, and it is larger at the start than at the stop, so the measured time is systematically short. Improve by using light gates, or by timing many oscillations and dividing. | 秒表：反应时间约 0.2 s，且启动时比停止时更大，因此测得时间系统性偏短。改用光电门，或测多个周期再相除。
- Metre rule: resolution 1 mm, and parallax when the eye is not level with the mark. Improve by using a vernier scale or travelling microscope, and by viewing perpendicular to the scale. | 米尺：分辨率 1 mm，且视线未与刻度齐平会产生视差。改用游标尺或读数显微镜，并垂直于刻度读数。
- Micrometer: zero error, and over-tightening the ratchet compresses a soft wire. Improve by recording and subtracting the zero reading, and by using the ratchet consistently. | 千分尺：存在零点误差，且棘轮拧得过紧会压扁软金属丝。记录并扣除零点读数，并统一使用棘轮。
- Thermometer: slow response, so a reading lags behind the true temperature; the stem may not be fully immersed. Improve by stirring and waiting for a steady reading, and by checking the immersion depth. | 温度计：响应慢，读数滞后于真实温度；玻璃管可能未完全浸入。搅拌并等待读数稳定，同时检查浸入深度。
- Balance: reading drifts with air currents or a warm object. Improve by shielding from draughts and by allowing the sample to cool to room temperature. | 天平：气流或温热物体会使读数漂移。加挡风罩，并让样品冷却至室温。
- Protractor and ray box: the ray has finite width and the protractor centre may not sit on the ray. Improve by projecting the ray onto paper and marking its centre line, and by taking the mean of both sides of the normal. | 量角器与光具座：光线有一定宽度，量角器圆心可能未落在光线上。把光线投射到纸上并标出中心线，并在法线两侧各测一次取平均。
5. Timing and Dynamics Experiments | 计时与动力学实验
For a simple pendulum, the standard relationship is T = 2π√L ÷ g enuine limitations are that the amplitude may exceed a few degrees so the motion is no longer simple harmonic, that the length is measured to the centre of the bob rather than the top of the string, and that the period is short enough for reaction time to matter.
对于单摆，标准关系式为 T = 2π√L ÷ g，变形后得 g = 4π²L ÷ T²。真正的局限性在于：摆角可能超过几度，运动不再是简谐运动；长度是量到摆球中心而非悬点；以及周期太短，反应时间会显著影响结果。
g = 4π²L ÷ T²
A strong pair of evaluation points reads: 'The angle of swing was not controlled and may have exceeded 10°, so the period was slightly longer than the small-angle value; use a protractor to keep the amplitude below 5°.' And: 'Only 10 oscillations were timed, so a 0.2 s reaction-time error is 2% of the measured period; time 50 oscillations or use a light gate at the equilibrium position.'
一组有力的评价点可以写成：'摆角未加控制，可能超过 10°，因此周期略大于小角度时的数值；用角规把振幅控制在 5° 以内。' 以及：'只测了 10 个周期，0.2 s 的反应时间误差占测得周期的 2%；改测 50 个周期，或在平衡位置用光电门计时。'
g is the dominant uncertainty, and because the fall is short the percentage uncertainty in t is large. Release mechanisms that drag on the ball, and the difficulty of judging the exact instant of impact, are both worth stating, together with the light-gate or trapdoor improvement.
对于自由落体，由 s = ut + ½at² 且 u = 0 得 a = 2s ÷ t²。由于 t 是平方项，计时是主要的不确定度来源，而下落时间很短，t 的百分比不确定度很大。释放机构对球的拖拽、以及难以判断准确撞击瞬间，都值得写出，并配上光电门或活门计时器的改进。
6. Electrical Experiments | 电学实验
enuine limitations: the wire heats up as current flows so its resistance drifts upwards during a reading, the diameter measured by micrometer varies along the wire, and the crocodile clips have contact resistance that adds to the measured value.
测电阻率时 ρ = RA ÷ L，而 A = πd² ÷ 4，因此工作公式为 ρ = πd²R ÷ (4L)。这里能写出很多真实局限性：通电时金属丝发热导致读数过程中电阻不断上升；千分尺测得的直径沿丝长并不均匀；鳄鱼夹的接触电阻会叠加到测量值上。
ρ = πd²R ÷ (4L)
- Heating: keep the current small, switch off between readings, and take the reading quickly. | 发热：电流取小值，读数之间断开电源，并快速读数。
- Uneven diameter: measure d at five positions along the wire and use the mean. | 直径不均：沿丝测五个位置的 d，取平均。
- Contact resistance: use a four-terminal arrangement, or subtract the resistance measured with the clips touching each other. | 接触电阻：采用四端接法，或先让两夹子直接接触测出接触电阻再扣除。
- Voltmeter loading: a real voltmeter has finite resistance and draws current, so it slightly reduces the reading. Use a high-resistance digital voltmeter. | 电压表分流：真实电压表电阻有限会分流，使读数略降。改用高阻数字电压表。
nal p.d. against current has gradient −r and intercept ε. The limitation is that the cell warms and its internal resistance falls during the run, and that the ammeter itself adds resistance to the external circuit. A good improvement is to take readings quickly, allow the cell to cool between runs, and use a low-resistance ammeter.
测内阻时 ε = I(R + r)，写成 V = ε − Ir，因此路端电压对电流作图，斜率是 −r，截距是 ε。其局限性在于：测量过程中电池发热使内阻下降，且电流表本身给外电路增加了电阻。好的改进是快速读取数据、两组之间让电池冷却、并使用低阻电流表。
7. Thermal and Optical Experiments | 热学与光学实验
measured temperature rise too small and therefore makes c appear too large. Simply writing 'use insulation' is thin; the stronger answer names the mechanism, then the method: ' ing temperature against time and reading the gradient at t = 0 to remove the cooling correction.'
在比热容实验中，c = IVt ÷ (mΔθ)。主要局限性是向环境散热，使测得的温升偏小，因而 c 偏大。只写 '加保温' 太单薄；更强的答案先点明机理再给出方法：'整个加热过程中金属块向空气散热，导致 Δθ 偏小；应给金属块加保温层，并采用初始速率法，作温度–时间图并读取 t = 0 处的斜率以扣除冷却修正。'
c = IVt ÷ (mΔθ) L = IVt ÷ m
oling problem appears, plus the difficulty of knowing exactly when all the ice has melted and the fact that the melted water may not be stirred, so the thermometer reads a temperature that is not representative.
测潜热时 L = IVt ÷ m，同样存在散热问题，此外还有冰是否恰好完全熔化难以判断，以及熔化的水未搅拌导致温度计读数不具代表性。
h of 1/v against 1/u. Limitations include the finite thickness of the lens so distances are measured to the lens holder rather than the optical centre, the difficulty of deciding when the image is exactly in focus, and the fact that a screen has to be moved back and forth to find the sharpest image. Improvements: measure from a fixed reference mark, repeat the focus judgement several times and average, and use a distant object to obtain f directly as a cross-check.
光学中常用薄透镜公式 1/f = 1/u + 1/v，通过作 1/v 对 1/u 的图像求 f。局限性包括：透镜有厚度，距离是量到透镜支架而非光心；难以判断像何时恰好清晰；必须前后移动光屏才能找到最清晰的像。改进：从固定参考标记量起，多次重复判断清晰位置并取平均，并用远处物体直接成像来交叉验证 f。
nλ = d sinθ v = fλ
at the grating spacing d may be quoted with limited precision, that the angle is measured between two first-order maxima so any misalignment of the laser doubles, and that the maxima are broad. Improvements: measure the angle on both sides of the zero order and halve the difference, and use higher orders to reduce the relative angular uncertainty.
在衍射光栅实验中 nλ = d sinθ，其局限性是：光栅常数 d 给出的精度有限；角度是在两个一级主极大之间测量的，因此激光的任何偏斜都会被放大一倍；主极大本身较宽。改进：在零级两侧各测一次角度并取差值的一半，同时使用更高级次以降低相对角度不确定度。
8. Writing the Improvement: Match, Mechanism, Magnitude | 写出改进：一一对应、作用机理、量级
A full-mark improvement has three components. First, it must match the limitation exactly, using the same apparatus vocabulary. Second, it must explain the mechanism by which the error is reduced. Third, where possible it should give a magnitude, so the examiner can see that you know the improvement is worth the effort.
满分的改进包含三个要素。第一，必须与局限性严格对应，使用同一套仪器词汇。第二，必须说明误差被减小的机理。第三，尽可能给出量级，让考官看出你知道这个改进值得做。
| Weak | Strong |
| Use better equipment. | Replace the hand-held stopwatch with two light gates connected to a data logger, removing the 0.2 s reaction-time error entirely. |
| Be more careful. | View the vernier scale perpendicular to its plane to remove parallax, which could otherwise contribute up to 0.5 mm. |
| Repeat the experiment. | Take five readings of the extension at each load and use the mean, reducing the random uncertainty by a factor of about √5. |
The right-hand column is longer, but it is also the version that earns marks, and it is not significantly harder to write once the habit is formed. Notice that each strong answer names an instrument, an action, and a numerical consequence.
右栏文字更长，但它才是得分的版本，而且一旦形成习惯，写起来并不明显更难。注意每个强答案都点出了仪器、动作和数值后果。
9. Extending the Range and Improving the Graph | 扩大测量范围与改善图像
Some evaluation marks are not about error at all, but about the quality of the data set. If all the readings lie in a narrow band, the gradient and intercept of the best-fit line are poorly determined, and any quantity derived from them inherits a large uncertainty. Stating this, and proposing a concrete extension, is a legitimate and often neglected evaluation point.
有些评价分与误差无关，而与数据集的质量有关。如果所有读数都落在很窄的区间内，最佳拟合直线的斜率和截距就确定得很差，由它们推出的任何物理量都会继承很大的不确定度。写出这一点并提出具体的扩展方案，是一条正当且常被忽视的评价点。
- Take readings over a wider range of the independent variable. | 在自变量的更宽范围内取点。
- Take more readings in the region where the graph is most curved. | 在图像弯曲最明显的区域增加读数。
- Repeat every point at least twice and plot the mean with an error bar. | 每个点至少重复两次，并画出带误差棒的平均值。
- Check that the first reading is not taken before the system has settled. | 检查第一个读数不是在系统尚未稳定时读取的。
For example, in a Hooke' s from 1.0 N to 1.5 N are used the extensions may all be under 5 mm, which is only a few divisions on a millimetre scale. Extending the range to 5.0 N gives extensions up to several centimetres and a much better-determined gradient, provided the elastic limit is not exceeded.
例如在胡克定律实验中 k = mg ÷ x，如果载荷只在 1.0 N 到 1.5 N 之间，伸长量可能都不到 5 mm，在毫米刻度上只有几格。把范围扩展到 5.0 N，伸长量可达数厘米，斜率就确定得好得多，前提是不超过弹性限度。
10. The Language of Evaluation | 评价题的实用句式
Evaluation answers are marked quickly, so clear sentence frames help the examiner find the marks. The three most reliable frames are the ones that build in the cause, the effect and the remedy automatically.
评价题的批改速度很快，因此清晰的句式能帮助考官迅速找到得分点。最可靠的三类句式会自动包含原因、结果和补救措施。
- Cause and effect: 'Because ..., the measured value of ... was systematically too ... by about ...'. | 因果句：'由于……，……的测量值系统性偏……，约……'。
- Remedy: 'This could be reduced by ..., which would lower the uncertainty in ... to about ...'. | 补救句：'这可以通过……来减小，从而把……的不确定度降到约……'。
- Range: 'The readings only covered ... to ..., so the gradient is poorly determined; extending the range to ... would ...'. | 范围句：'读数只覆盖……到……，因此斜率确定得很差；把范围扩展到……将……'。
Write each point as a separate short paragraph or bullet. Do not bury an evaluation point inside a long description of the method, and never write 'human error', which tells the examiner nothing about the procedure.
把每个点写成独立的短段或要点。不要把评价点埋在冗长的方法描述里，也绝不要写 '人为误差'，这句话对考官毫无信息量。
11. Two Worked Model Answers | 两篇范例答案
Model answer one, a pendulum used to determine g. Limitation: 'The string was fixed to a clamp on a retort stand and the stand could wobble slightly, and the pivot was a knot rather than a sharp support, so the effective length was not well defined; the length was also measured to the top of the bob rather than its centre.' Improvement: 'Use a split cork in a boss and clamp so the pivot is fixed and sharp, measure L from the pivot to the centre of the bob with a metre rule viewed perpendicular to the scale, and take the mean of several readings.'
范例一：用单摆测 g。局限性：'摆线固定在铁架台的夹子上，铁架台会轻微晃动，而支点是绳结而非尖锐支撑，因此有效摆长定义不明确；此外长度是量到摆球顶端而不是球心。' 改进：'用夹在十字夹中的开缝软木塞做成固定而尖锐的支点，用米尺从支点量到摆球中心，读数时视线垂直于刻度，并取多次读数的平均。'
Model answer two, an electrical experiment to find the e.m.f. and internal resistance of a cell. Limitation: 'Readings were taken with a digital voltmeter across the cell while the current changed from 0.10 A to 0.60 A, and the cell warmed noticeably during the run, so its
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图