Download the templates
A blank CSV, a worked CSV with the paper reference and attribution, and a TXT worksheet with the complete reading record. The example concerns a real published study; the prompts and commentary are original. The files contain no participant-level data, automatic analysis or copied university worksheet.
- Blank one-paper reading recordCSV · 187 bytes
- Eight checked reading records for Endres and Renkl (2015)CSV · 5 KB
- Text worksheet with the complete worked reading record and attributionTXT · 6 KB
Open a blank file for your own work. Keep examples separate from your evidence and follow the instructions below.
On this page
Use the worksheet to inspect one empirical article
A research article reading worksheet helps you recover what a study actually did before you summarize it. Write your reason for reading, identify the paper and record whether you inspected the full text. A title and abstract can help you decide relevance; they may not contain the design, scoring or analysis needed for your intended claim.
This record is intended for learning to read an empirical study, especially an experiment with Methods and Results sections. Adapt it for other designs and disciplines. A conceptual essay or qualitative study may require different fields; an absent numerical result is not automatically a defect.
Tulane distinguishes the roles of an article’s sections; Stanford connects a reading goal with structural inspection and explaining a study in your own words. Use those ideas to choose where to look, rather than assume every paper follows one rigid reading order. The worksheet and example here are our own materials.
Match each reading question to a source location
Skim the headings and identify the authors’ question, then trace the relevant design and result. The Introduction usually supplies context and an aim; Methods describes how the work was done; Results reports analyses; Discussion interprets them. Check the actual article layout instead of filling a generic template from memory.
Record a section heading together with a table, figure or distinctive paragraph when useful. A publisher page can reflow across devices, so a screen position is a poor locator. In a PDF, keep the printed page number and the table or subsection as well. Preserve the article reference and version with your notes.
Separate three states: not checked, checked but unclear, and not reported in the inspected material. If you cannot find an exclusion count, that does not establish that there were no exclusions. State what you inspected and leave the remaining question visible. Do not manufacture a complete sample-flow diagram from one reported N.
Follow eight checked fields in a real study
Endres, T., & Renkl, A. (2015). Mechanisms behind the testing effect: an empirical investigation of retrieval practice in meaningful learning. Frontiers in Psychology, 6, Article 1054. https://doi.org/10.3389/fpsyg.2015.01054
The example uses the full publisher article, not an abstract-only summary. It examines retrieval practice with university students and three texts. It does not test an AI service, provide a complete review of study techniques or report a new experiment by AI Essay.
Worked notes paraphrase Endres and Renkl (2015), © 2015 the authors, published in Frontiers in Psychology under CC BY 4.0: https://creativecommons.org/licenses/by/4.0/. The reading prompts and teaching commentary are AI Essay editorial additions; no author endorsement is implied.
Open the paper beside this record. Before reading a completed entry, locate its source section and write your own answer. The last column explains a reading boundary or next action; it is editorial teaching commentary, not an additional finding reported by the study.
Scroll the table sideways to view all columns.
| Reading question | Checked record | Source location | Boundary or next check |
|---|---|---|---|
| What did this study ask? | Compare free-recall and short-answer retrieval with restudy, and examine proposed explanations of the testing effect with structured texts. [2] | Introduction, final paragraphs before Materials and Methods | Keep the study aim separate from your own essay question. This experiment does not evaluate an AI study service. |
| Who is represented by N? | Subjects reports 47 university students. Table 2 repeats N = 47 for each condition; the earlier recruitment total and a separate dropout flow are not supplied. [2] | Materials and Methods → Subjects; Results → Main Analysis → Table 2 | Do not add the repeated condition counts to produce 141 participants or invent a recruitment-to-completion history. |
| What differed between conditions? | The same participants experienced restudy, free recall and short-answer practice on different texts. Text–condition combinations and presentation order were randomized. [2] | Materials and Methods → Design; Procedure; Figure 1 | Within-participant conditions are not three independent participant groups. Inspect how each text was assigned before interpreting a condition difference. |
| What was measured, and when? | Three psychology texts translated into German were used. Two computer-based sessions were one week apart; the second assessed all texts using free recall followed by short answers. [2] | Materials and Methods → Materials; Procedure, second-session paragraph; Figure 1 | Practice format and later assessment format are separate fields. A one-week text assessment is not a course grade or an indefinite retention measure. |
| What units do the table numbers use? | Conceptual aspects were coded, then text-specific outcomes were z-standardized. Table 2 short-answer assessment means are −0.24 for restudy, −0.04 for free-recall practice and 0.29 for short-answer practice. [2] | Materials and Methods → Materials, coding paragraph; Results → Pre-analyses; Table 2 | These are standardized outcomes, not percentages or raw marks. The value 0.29 is not 29% correct; do not calculate a percentage improvement from these values. |
| Which comparison supports the main result? | The combined retrieval-versus-restudy contrast was significant, while the overall comparison of the two retrieval formats was not significant (p = .118). [2] | Results → Main Analysis, paragraphs around Table 2 | Keep the actual contrast attached. A combined contrast does not establish every separate pairwise difference, and non-significance does not establish equivalence. |
| Does a result concern a subset of questions? | For initially non-tested short-answer items, short-answer practice exceeded restudy. The exploratory comparison with free-recall practice did not detect a difference. [2] | Results → Main Analysis, paragraphs around Table 3; Table 3 | These questions belong to the studied texts. Do not turn a result for this item subset into proof of transfer to every new topic or task. |
| What is interpretation rather than an independently tested mechanism? | Participants rated mental effort on a 0–100 scale. Condition differences disappeared after statistical adjustment, which the authors interpreted as support for elaborative retrieval. [2] | Materials and Methods → Materials, effort rating; Results → Main Analysis, adjustment paragraph; Discussion → Review of Hypothesis | Editorial inference: effort was measured rather than independently randomized. Adjustment alone does not independently establish the proposed causal mechanism. |
Read the table with its design, units and contrast
Start with the denominator. The same 47 participants contributed to the three practice conditions. Repeated N values in Table 2 do not represent three separate samples. The Methods design makes the count interpretable; reading the table alone invites a different story.
Next inspect the units. Table 2’s short-answer assessment means include −0.24 and 0.29, after text-specific z-standardization. These values cannot be read as percentages correct. The first is not negative knowledge and the second is not 29% of a test. Keep the assessment label and the scoring procedure beside the numbers.
Then identify the comparison. A result combining the retrieval conditions against restudy is a different contrast from comparing free-recall practice directly with short-answer practice. The non-significant overall format comparison does not show that the formats are interchangeable. The item subset in Table 3 needs its own description.
Finally separate explanation from observation. The authors interpret the mental-effort analysis as support for an account of retrieval practice. Our design-based caution is that self-rated effort was measured, not independently randomized. Recording the authors’ explanation does not require presenting it as a separately proven causal mechanism.
Supporting reading: [2]
Attempt three checks before opening the worked answers
Use the linked article and make a note before revealing each answer. These are reading exercises with editorial explanations, not an automated assessment or a product accuracy test. There is no score and nothing is submitted or saved on this page.
Table 2 lists N = 47 in each condition. Did the experiment have 141 different participants?
No. The Subjects section reports 47 students and the Design describes a within-participant comparison. The same people appear across the condition rows. Do not add those repeated counts as if they were independent groups. [2]
Can the short-answer assessment mean of 0.29 be reported as 29% correct?
No. The Pre-analyses section describes z-standardization. Table 2’s mean is expressed in those standardized units. A percentage requires the relevant raw scoring information and denominator, not multiplying this value by 100. [2]
Do the adjusted mental-effort results independently prove why retrieval practice worked?
Our reading is more limited. The authors interpret the pattern as support for elaborative retrieval, but the effort measure is self-reported and was not independently randomized. Keep that explanation separate from the condition comparison actually tested. [2]
Write a conclusion whose scope the record can support
A bounded note is: in this experiment with 47 university students and three texts, the combined retrieval conditions exceeded restudy on a one-week assessment; the overall comparison between the retrieval formats was not significant. This preserves the context and the contrast without promising that every quiz format improves every learning outcome.
Before using the result in your work, inspect whether that context answers your question. A school-age population, a different subject or a course-grade outcome would require additional evidence. The sample, materials and delay are boundaries of this example, not a claim that the paper reported every possible limitation.
If a summary changes repeated condition rows into independent participant groups, a standardized mean into a percentage or an interpretation into a proven mechanism, stop at that exact field and reopen its source. The same habit applies when a human or an AI tool suggests the reading record. A source link is a route back to evidence, not a completed check.
Supporting reading: [2]
Move checked reading notes into the next writing task
Use the blank CSV for a different paper. Keep the reference in your own file or first rows, adapt the questions and retain exact source locations. The worked CSV includes reference and attribution rows above the four-column table; locate its header if importing into another sheet. The TXT version preserves the instructions and full example in one readable file.
An annotated bibliography needs a concise summary, evaluation and relevance to your project. Choose those points from the checked record rather than paste every worksheet field. A literature matrix serves a later comparison across papers. This study appears as S2 in our existing matrix so you can trace the handoff deliberately.
A reading assistant may help identify passages to inspect, if its use and the material are permitted. Keep its proposed field values separate until you check them. The manual workflow here works without an account or an AI service; it provides no finding about the accuracy of a paper-summary tool.
Questions about this topic
How is this different from an annotated bibliography or literature matrix?
This worksheet teaches how to recover a single paper’s question, design, measurement and result from the original sections. An annotated bibliography prepares a source entry; a matrix compares sources. Use the reading record as working evidence for those later tasks.
Do I need the full text?
The worked example uses the full article, including Methods, Results and tables. If you have only an abstract or part of a paper, record that access limit and leave fields unresolved. Do not imply that you inspected details unavailable in the material you read.
Is the study example real?
Yes. The worked record paraphrases Endres and Renkl’s 2015 open-access article with attribution and source locations. The worksheet prompts and commentary are editorial additions. We did not run the experiment or a test of an AI product.
Does completing the worksheet prove a paper is reliable?
No. A filled record makes your reading inspectable. It does not replace disciplinary appraisal, instructor feedback, further source checks or a systematic review, and it supplies no universal credibility score.
Sources and editorial notes
The supporting reading below informs our editorial guidance. Exercises and scenarios illustrate a method; they do not report a completed product test or measured learning outcome. Read our editorial policy or send a correction with a supporting source to info@ai-essay.com.