LET Secondary Assessment of Learning — Constructing, Administering and Analyzing TestsSummary

The stem is the question or problem posed in an MCQ. It must state a single, clear problem so that learners understand what is being asked before they read the options. Effective stems place as much wording as possible in the stem itself, keeping the options brief. They avoid negatives unless absolutely necessary (and if a negative such as NOT or EXCEPT must appear, it should be emphasized in bold or capitals). For example, in a Grade 5 Science MCQ about photosynthesis, a weak stem is "Photosynthesis is..."; a strong stem is "Which of the following best describes the process by which plants convert sunlight into chemical energy?" The latter makes the question crystal clear. Avoid unintended clues, such as grammatical give-aways (the use of "a" or "an" that points to a specific option) or vocabulary mismatches between the stem and an incorrect option.

An MCQ has one correct answer (the key) and several incorrect answers (distractors). Good options are plausible and reflect common misconceptions or errors that learners might make. For instance, in a Grade 3 Mathematics MCQ about addition, a correct answer might be 15, while distractors could be 13 (if the student miscounts) or 14 (if the student forgets to carry over). Options must be homogeneous in content, grammar, and length; the correct answer should not stand out visually by being the longest or most detailed. All options should be grammatically parallel and reasonable in scope. Avoid absolutes (such as always, never) in distractors—these are red flags that savvy students eliminate immediately. Use "all of the above" and "none of the above" sparingly; "all of the above" allows a student who knows only two options to answer correctly by elimination, defeating the purpose of the distractor.

True-false items state a claim and ask learners to judge its truth. They should test a single, clear idea with a statement that is unambiguously true or unambiguously false. In a Grade 2 Health MCU on nutrition, "Vegetables are good for your body" is true; "Sugar is the only source of energy" is false. Avoid specific determiners—words like all, always, never, and none that statistically signal false statements, while sometimes, usually, and generally signal true statements. Skilled test-takers can use these linguistic patterns to guess correctly without knowing the content. Also avoid double negatives ("It is not true that plants do not need water") and trivial tricks. Keep roughly equal numbers of true and false statements and vary their order randomly throughout the test.

Matching items present a list of premises (questions or stems) on the left and a list of responses (answers) on the right. Students draw lines or write letters to pair them. Effective matching sets are homogeneous—all premises and responses address the same topic or principle, with one clear basis for matching. For instance, matching Grade 4 learners might match Philippine historical figures on the left (Jose Rizal, Andres Bonifacio, Emilio Aguinaldo) with their key accomplishments on the right. Always provide more responses than premises (if there are 5 premises, provide at least 7 responses) so the last match cannot be solved by elimination alone. Keep lists short (5 to 8 premises) for readability and manageability. Place the shorter list on the right (responses) so the eyes scan efficiently. Give clear directions stating the basis for matching and whether responses may be used more than once.

Completion items present an incomplete statement and require learners to fill in one brief, correct answer. For example, "The three branches of the Philippine government are the legislative, executive, and _____" (answer: judicial). The blank should be placed near the end of the statement, after the problem is clear, not at the beginning or middle where it disrupts comprehension. Use blanks of equal length so that the physical length of the blank is not a clue; avoid grammatical clues such as placing "a" or "an" before the blank, which telegraphs the type of answer (consonant or vowel sound). In Grade 5 Science, ask "The gas that plants release during photosynthesis is _____" (answer: oxygen), not "The gas that plants release during photosynthesis is a _____" or "an _____," which hints at the answer.

Restricted-response essays limit the content and form of the answer by posing a narrow, focused task. Examples include "List three causes of the 1896 Philippine Revolution and explain each in one or two sentences," or "Compare and contrast the characters of Sisa and Kabesang Tales in Noli Me Tangere in three sentences." Because the scope is defined, learners spend less time deciding what to include and more time demonstrating knowledge. Restricted-response essays are easier to score reliably because the expected content is clear, and they allow you to sample more content (you can ask five restricted-response items more efficiently than one extended-response item). They are ideal for assessing whether learners can organize and explain information in a structured way.

Extended-response essays give learners freedom to organize, argue, and demonstrate higher-order thinking. An example is "Evaluate how the K-12 Senior High School reform has changed the way teachers prepare Grade 6 learners for future careers. Support your answer with specific examples." Learners choose what to include, how to structure their argument, and how deeply to develop each point. Extended-response essays are rich assessments that tap synthesis, evaluation, and creativity, but they are harder to score reliably because different students may take different valid approaches. Without a clear rubric, two teachers might score the same essay very differently. Extended-response essays are most appropriate in upper elementary grades (Grades 5–6) where learners have developed sufficient writing proficiency.

Before administering an essay test, prepare a model answer (or a detailed rubric) that defines the criteria for scoring. For a restricted-response item like "List three causes of World War II," the model answer might specify the three causes to accept (e.g., the Treaty of Versailles, economic depression, rise of fascism) and state that one sentence of explanation per cause is required. For an extended-response item, develop an analytic rubric that breaks scoring into components (e.g., thesis statement, evidence, organization, grammar, each worth a certain number of points) or a holistic rubric that assigns overall quality levels (Excellent, Good, Fair, Poor). Use the point method: score one question across all papers before moving to the next, rather than scoring all questions on one paper and then moving to the next paper. This method keeps a consistent standard and prevents fatigue-related inconsistency. Where possible, score essays anonymously to reduce halo bias—the tendency to rate a learner higher because of a previous impression (positive or negative).

After you have written all your items, assemble them into a coherent test. Group items by type: place all true-false items together, all MCQs together, and so on. This organization helps learners shift mental strategies smoothly and makes administration and scoring easier. Arrange items from easy to difficult so that learners build confidence early and are not discouraged by hard items at the start. (A Grade 3 arithmetic test might begin with simple addition and progress to word problems.) Write clear directions for each section, specifying how many points each item is worth and what learners should do (e.g., "Circle the letter of the best answer" or "Write your answer on the line provided"). Keep each item and its options on the same page so learners do not have to flip back and forth. Space items generously for readability and to prevent copying. Prepare the answer key and scoring plan (e.g., 1 point per MCQ, 2 points per essay) before printing the test. This preparation prevents errors and clarifies how you will grade.

The setting in which you administer a test affects its validity. Provide a comfortable, well-lit, quiet space free of distractions. In a crowded Philippine classroom, this might mean asking students to move desks apart to prevent cheating and to give each student adequate space. Give clear oral and written instructions at the start: read the directions aloud, point to sections, and state the time limit unambiguously. For example: "You have 45 minutes to complete this test. Stop writing when I say time. Raise your hand if you have questions." Minimize test anxiety by reassuring learners that the test is an opportunity to show what they know, not a trick. Discourage cheating through careful seating arrangements and active proctoring; walk around the room, maintain an authoritative but calm presence, and watch for suspicious behavior without accusation. Manage timing fairly so that all learners have a reasonable chance to finish; do not penalize slow workers simply for pacing if they have genuinely engaged with the task. For younger learners (Grades 1–3), consider breaking tests into shorter sessions to respect attention spans.

Most classroom objective tests are scored simply by counting correct answers. However, on some high-stakes assessments or when a test has few items and high guessing rates, a correction for guessing may be applied. The formula is: Score = R - W / (k - 1), where R is the number of right answers, W is the number of wrong answers, and k is the number of options per item. Omitted items (blanks) are not counted as wrong. The logic is that on a pure guess, a learner has a 1 in k chance of being right; therefore, if they guess on three items with 4 options each, they expect to guess 1 right and 2 wrong. The formula "refunds" that lucky guess: Score = (1 + 2 + actual rights) - 2 / (4 - 1) accounts for the guessing. Example: a student answers 40 items correctly, misses 9, and omits 1 on a 50-item, 4-option test. Corrected Score = 40 - 9 / (4 - 1) = 40 - 3 = 37. The correction lowers the score, which is harsh on careless guessing but generous on educated guesses (a student who is 75% sure and guesses is penalized lightly). Most DepEd classroom tests do not use the correction, but the LET has tested both the computation and its rationale, so be familiar with it.

The difficulty index, denoted p, is the proportion of examinees who answered an item correctly. It is computed as: p = (number of correct responses) / (total number of examinees). The index ranges from 0.00 to 1.00. Counterintuitively, a higher p means an easier item, because more people got it right. A lower p means a harder item. For example, if 30 of 40 students in a Grade 5 class answered an MCQ correctly, p = 30 / 40 = 0.75, indicating an easy item. If only 8 of 40 answered it correctly, p = 8 / 40 = 0.20, indicating a very difficult item. The interpretation bands are: 0.00–0.20 (very difficult), 0.21–0.40 (difficult), 0.41–0.60 (moderately difficult / average), 0.61–0.80 (easy), and 0.81–1.00 (very easy). An ideal difficulty index is around 0.50, where the item best separates strong from weak learners. (For 4-option multiple-choice, some sources cite an ideal near 0.60 to account for the 0.25 guessing baseline.) Items that are too easy (p above 0.85) or too difficult (p below 0.15) contribute little discrimination power and are candidates for revision or removal.

The discrimination index, denoted D, measures how well an item separates high-scoring learners from low-scoring learners. It is computed from the upper group (typically the top 27% of students, or the upper half) and the lower group (the bottom 27%, or the lower half). The 27% convention comes from Kelley's research and maximizes contrast while keeping group sizes stable. In a class of 40, using 27%, each group would contain about 11 students (0.27 × 40 = 10.8, rounded); in practice, many teachers use the upper and lower halves (top 20 and bottom 20 of a 40-student class) for simplicity. The formula is: D = (number correct in upper group - number correct in lower group) / (number of students in one group). For example, if 8 of 10 upper-group students answer an item correctly and 3 of 10 lower-group students do, then D = (8 - 3) / 10 = 0.50. The index ranges from -1.00 to +1.00. A positive D indicates the item favors the upper group (good), meaning stronger learners are more likely to answer correctly. A negative D indicates the lower group outperformed the upper group, a red flag that the item is flawed (e.g., miskeyed, ambiguous, or testing a concept the upper group misunderstood). Interpretation: D of 0.40 and above is excellent; 0.30–0.39 is good; 0.20–0.29 is fair but needs improvement; 0.19 and below is poor; negative D is defective.

Beyond computing overall difficulty and discrimination, examine how each distractor (incorrect option) performed. A good distractor attracts more lower-group than upper-group students; it is seductive because it reflects a plausible misconception. A non-functional distractor is chosen by few or no students across both groups; it does not trick anyone and wastes a slot. A distractor chosen by more upper-group than lower-group students signals an ambiguous or miskeyed item. For example, imagine an MCQ with key C. In the upper group of 10, option C is chosen by 8 students, option A by 1, option B by 0, and option D by 1. In the lower group of 10, option C is chosen by 3 students, option A by 3, option B by 2, and option D by 2. Option A and D function as distractors (attracting more lower-group students). Option B is non-functional (attracting only 2 lower-group students and no upper-group students). The key (C) is clearly chosen more by the upper group, yielding a positive D, which is healthy. However, if the pattern on C were reversed (3 upper, 8 lower), D would be -0.50, and the first step would be to verify that C is indeed the correct answer; if it is, the item or its options are flawed and need revision.

After computing difficulty and discrimination and analyzing distractors, decide what to do with each item. Retain items that have good difficulty (roughly p = 0.30 to 0.70) and discrimination of 0.30 or higher. These items are functioning well and reliably differentiate learners. Revise items with marginal discrimination (D = 0.20–0.29) or non-functioning distractors (e.g., an option chosen by no one); these items have potential but need improvement—perhaps the wording is unclear, a distractor is implausible, or the stem is misleading. Reject or completely rewrite items with very low discrimination (D below 0.20) or negative discrimination, and always verify the answer key when the upper group underperforms on the keyed response. A negative D almost always signals a miskeying error or a deeply flawed item. In a large item bank, weak items are discarded; in a smaller, ongoing classroom context, you may set aside the item and use it in a revised form next term. The decision should be data-driven and documented.

Here is a full example to integrate all concepts. You administer a Grade 4 Science test with 20 MCQs on the water cycle. You analyze the upper 10 and lower 10 students. Item 7 (with key C) yields this response pattern: Option A (upper: 1, lower: 3), Option B (upper: 0, lower: 2), Option C / Key (upper: 8, lower: 3), Option D (upper: 1, lower: 2). Difficulty (from both groups combined): p = (8 + 3) / 20 = 0.55 (moderately difficult, in the ideal band). Discrimination: D = (8 - 3) / 10 = 0.50 (very good / excellent). Distractor analysis: Option A attracts more lower-group students (3 vs. 1), so it functions. Option B attracts some lower-group students (2) but none from the upper group, so it is light but functional. Option D draws equal numbers (1 and 2), so it is weak. The key C is clearly favored by the upper group. Verdict: Retain the item as written. It is doing its job of separating high from low achievers.

The quality of individual items directly determines whether the test measures what it claims to measure. A test composed of poorly written items (ambiguous stems, implausible distractors, grammatical clues) yields invalid scores that do not reflect true learning. Conversely, items written to the guidelines—clear stems, homogeneous options, and functional distractors—create a test that validly assesses the intended objectives. This relationship is foundational to the entire assessment process.

How you assemble and administer a test directly affects whether the results are fair and trustworthy. A well-organized test (grouped by type, arranged easy to difficult, clearly formatted) is easier for learners to navigate and reduces confusing. Fair administration (quiet setting, clear instructions, adequate time, active proctoring) ensures that learners' scores reflect their knowledge, not test anxiety, cheating, or environmental chaos. Together, these practices uphold RA 7836's ethical requirement that assessment be fair and defensible. A test may be well-written, but poor administration undermines its validity.

Difficulty and discrimination work together to characterize an item's quality. An item with ideal difficulty (p ≈ 0.50) has the most room to discriminate, because about half the class got it right and half did not. A very easy item (p > 0.85) or very hard item (p < 0.15) has little room for the upper and lower groups to diverge, so D will be low even if the item is well-written. Thus, you want items that are moderately difficult and show good discrimination. An item with p = 0.50 and D = 0.40 is excellent; an item with p = 0.90 and D = 0.20 is easy but weak at differentiation.

A negative or very low discrimination index is an early warning sign to check the answer key. If the upper group (stronger learners) choose option A and the lower group (weaker learners) choose option B, but the key lists B as correct, the item will show negative discrimination. This is unnatural and suggests a miskeying error. Before revising or discarding the item, verify the key. If the key is correct and the pattern persists, the item is genuinely ambiguous or testing a concept the upper group misunderstood, and it should be revised or rejected. This relationship teaches a troubleshooting mindset: negative D is a symptom, not a diagnosis.

Distractor analysis reveals which parts of an MCQ are not working. A non-functional distractor (chosen by no one) should be replaced with a more plausible error. A distractor chosen primarily by upper-group students suggests the item is ambiguous or the key is questioned even by strong learners, prompting a rewrite. A distractor that attracts many more lower-group students is gold—it is doing its job of testing a specific misconception. Over multiple test cycles, distractor analysis data accumulates, and you can refine options to be increasingly powerful tools for identifying gaps in understanding.

The quality of your rubric or model answer directly affects the consistency and fairness of your essay scoring. A vague rubric ("Award points for good ideas") leaves you guessing and invites bias—you may score one learner's essay high because you like their handwriting or because they remind you of a good student (halo bias), and score another low for petty reasons. A detailed analytic rubric ("Thesis statement: 2 points; three supporting ideas with evidence: 3 points each; grammar and organization: 2 points; total: 13 points") and a model answer ("The thesis should name the three causes of X and state a judgment...") make scoring objective and defensible. The rubric thus ensures that two teachers grading the same essay would award similar scores.

Item analysis results (difficulty and discrimination indices) offer insight into whether your curriculum and instruction have been effective. If many items have p < 0.20 (very difficult), it may indicate that the content was taught too quickly, learners were not ready, or the content was not emphasized enough. If many items have p > 0.85 (very easy), it may mean the content was over-taught, learners already knew it, or the test was not challenging enough. Similarly, items with low discrimination may point to poorly understood concepts or gaps in instruction that the upper group also struggles with. Thus, item analysis feeds back into instructional planning: the next unit or next year's curriculum can be adjusted based on what the data reveal.