How to read a confession obtained under interrogation

Before you begin

A written confession seems to offer exceptional access to someone from the past. It speaks in the first person, lists places and explains actions. That proximity may mislead us: the document we preserve results from a communicative situation, an act of writing and a history of transmission. Learning to read it requires attention to all three without erasing the person whose statement was recorded. This lesson proposes a reading that preserves the testimony's historical value and limits the claims we can construct from it.

Our reference case is the confessions attributed to Isobel Gowdie at Auldearn, Scotland, in 1662. We do not decide whether the supernatural events narrated actually occurred or reconstruct inaccessible inner experience. The documentary question is: what do we know of a claim appearing in a confession made before authorities and known through copies, editions or studies? Distinguish narrated event, recorded statement, recording act and later interpretation. Each needs a different argument.

By the end, you will be able to describe the chain leading from a statement to a publication; separate voices and modalities within a passage; compare sessions without counting repetitions as independent witnesses; and write conclusions that do not turn accusations into facts. You will also learn to explain a precise lack of knowledge. Writing that we do not know an outcome through the corpus consulted may be a sound documentary conclusion, provided you state the search's scope and avoid confusing it with a claim about every possible archive.

The first part develops the problem and a complete fictional case file. Exercises with answers, a project and reference tools follow. The final sections adapt methodological materials from other lessons in this library, identified in their introductions. Their inclusion allows this volume to be used independently; they are not new documents about Gowdie. The workshop scenes, names and sentences are expressly identified teaching inventions intended to practise attribution, dependence and limits of interpretation.

Work in three sessions. In the first, read through to the case file and draw its relationships. In the second, complete the exercises before looking at the answers. In the third, produce a research record and use the reference workbook to review your decisions. You do not need to memorise Scottish vocabulary or adopt an interpretation of the case. You need to show where each claim comes from, which intermediaries it contains and which datum might require you to modify it.

1. Four questions behind a first-person statement

The phrase ‘I was there’ contains a speaking position, but does not by itself guarantee the historical presence of the person speaking in the place mentioned. In a judicial source, we must ask who appears as the speaker, who records the statement, to whom it is addressed and what it seeks to establish. An edition can preserve the pronoun while removing the context of production. Readers then receive an impression of immediacy greater than the document allows us to justify.

Four questions should be distinguished. The first is textual: what does the passage before us claim? The second is documentary: how did that claim reach this written form? The third is historical: how does it relate to events, beliefs or practices beyond the text? The fourth is historiographical: how have other readers used that passage? Answering one well does not automatically resolve the others. We may transcribe a sentence precisely and still not know whether it narrates an action actually performed.

Imagine a fictional statement in which a woman says she gave a ribbon to a neighbour. The record constitutes evidence that this transfer was recorded as part of the statement. Proving the material transfer would require further arguments: independent testimony, a relevant contemporary note or a securely connected object. To study the ribbon’s meaning, we would also need context. The step from a preserved sentence to historical practice does not occur simply because the account contains names and details.

Nor does mediation turn everything into the scribe's invention. That would be the opposite error. A record may preserve significant information about vocabulary, personal relationships, forms of narration or institutional expectations. What changes is how we attribute it. Instead of writing ‘the person thought exactly this’, we may write ‘the surviving statement attributes this explanation to the person’. The latter sentence keeps the relationship between voice, writing and experience open, and allows new evidence to be incorporated without undoing the entire analysis.

A manageable unit

Work with small claims. ‘The accused confessed freely, described an ancient ritual and was executed’ combines at least three problems: conditions of production, the genealogy of a practice and judicial outcome. A source documenting the account may resolve neither of the other two. Divide the sentence into propositions and assign each its reference. If one part lacks support, do not hide it behind the citation supporting another.

Analysis begins with a five-field record: the formulation you are studying; the person to whom it is attributed; the medium consulted; the context preserved; and the proposed inference. Add a sixth field for what is missing. This final column prevents a provisional reading from becoming unconditional certainty when copied into another work. The discipline consists in preserving the distance between what you see and what you deduce, even when both steps seem obvious.

2. Isobel Gowdie: one case, several mediations

The starting corpus brings together four statements associated with Gowdie in Auldearn in 1662. Their narrative richness has encouraged very different readings. For this lesson, the baseline information is limited: a set of statements produced in a context of witchcraft accusation; a nineteenth-century publication by Robert Pitcairn; and modern studies examining how the texts were constructed. We will not turn a striking narrative into verification of supernatural phenomena or a complete biography of an inner life.

Pitcairn's 1833 edition and the 1662 record belong to different moments. The Museum of Witchcraft and Magic catalogue identifies a copy of the confessions from that publication. The reference allows a transmission route to be recognised, but does not replace material comparison of all its witnesses. When citing a reading obtained from a printed work, identify that work. When research refers to a manuscript, do not present its information as your own observation of the original. ^[1]

Emma Wilby devoted a monograph to these confessions. The publisher's presentation of the version available on JSTOR describes the recovery of records thought lost and the aim of distinguishing Gowdie's contribution from that of her interrogators. It also presents an interpretation concerning visionary traditions that must be attributed to the author. The presentation consulted mentions an authentication process; it does not allow this lesson independently to assert its subsequent result or an inspection of the manuscripts. ^[2]

This caution is productive. A modern reference may correct the idea that only an old printed work is available to us, but the correction needs a date and provenance. The history of access to a source forms part of research. An author working before a documentary recovery did not necessarily have the same opportunities as a later author. Comparing their conclusions also requires comparing the available corpus, the questions and the reading techniques.

What remains open

The judicial outcome will not be established here by conjecture. That other accused people in comparable contexts suffered particular penalties does not demonstrate Gowdie's individual fate. A contextual probability is not equivalent to a sentencing document or execution record. The responsible formulation is local: the assemblage examined in this lesson does not establish an outcome. Progress would require identifying specific evidence and checking that it concerns the same person and procedure.

Nor will we choose between a completely spontaneous voice and total fabrication by the authorities. These extremes are too rigid to guide an initial reading. Our work will be to recognise operations: attributing, questioning, recording, summarising, ratifying, editing and commenting. Some may be documented; others may only be raised as alternatives. The result may explain how a statement circulated even when we cannot yet assign every word to a historical speaker with complete certainty.

3. From speech to a readable document

An oral statement unfolds over time. It can contain interruptions, repetitions, silences, gestures and responses to interventions that are not written down. A written record organises information on a durable medium. It can summarise, order and employ formulas familiar to those handling the proceedings. An edition, in turn, offers that text to other readers through decisions about transcription, punctuation, layout and commentary. This route should be described not as an automatic loss of truth, but as a sequence of operations that we must identify.

Separate a contemporary record from a later copy. A copy may carefully preserve a document now lost, while an original may be difficult to read or incomplete. ‘Original’ and ‘reliable’ are not absolute synonyms. The advantage of an early witness depends on the question: it may allow a correction or hand to be studied; a critical edition may provide variants a single image does not show. Describe which resource you use and for what purpose, rather than ranking every support on a universal scale.

Punctuation deserves attention. A full stop may turn a continuation into a self-contained answer. Modern quotation marks may suggest verbatim speech where the manuscript does not mark it. A paragraph may separate two formerly connected episodes or bring together statements from different moments. None of these possibilities authorises accusing an editor of manipulation. They are questions for checking: consult the criteria and, where relevant, compare the presentation with the witness it declares it follows.

An intellectual chain of custody

For each passage, record a minimal route: attributed statement; document containing it; witness or edition consulted; translation used; commentary that guided your reading. You do not need to resolve the entire textual tradition before beginning. Avoiding invisible jumps is enough. If you have read only a translation quoted by an article, your access passes through those two intermediaries. You can use it proportionately, but cannot describe it as a direct transcription of the case file.

A good note distinguishes checking from referral. ‘I checked this statement on the page indicated’ describes your work. ‘The study refers to another witness’ describes someone else's work. Both are useful and may coexist. The problem is merging them into a formula that makes readers believe every link has been reviewed. Traceability does not require omniscience; it requires an honest, sufficiently precise description of the route followed.

4. Seeking the question without inventing it

In an interrogation, answers may survive in greater detail than questions. A statement that suddenly lists names, specifies a date or explains a relationship offers clues to interaction. However, a clue is not a transcription of the lost question. The list could also reflect the speaker’s initiative or a reorganisation of the record. The task is to describe the change and formulate alternatives, rather than write an imaginary dialogue as though it belonged to the archive.

A simple method uses three levels. First, mark an observation: ‘the text moves from an account of travel to a list of participants’. Then formulate a hypothesis: ‘the change is compatible with a request for identification’. Finally, establish a limit: ‘the specific question is not preserved in the passage consulted’. Readers can thus see which part they can check and which part depends on inference. A newly found question record can test this without concealing the earlier reading’s provisional status.

The content of a question also matters. ‘What happened next?’ opens a different space from ‘who accompanied you?’ The latter presupposes accompaniment. An invented exercise can compare both formulations and explain their possible effects. A historical case cannot be assigned either one without evidence. This distinction allows us to teach how an interaction guides a narrative without turning an analytical possibility into something that happened to a particular person.

Traces and alternatives

Formulas of resumption, ratification or clarification may indicate stages in the procedure. Study their verbs and subjects. ‘It was read to her’ indicates a different operation from ‘she added’. ‘She ratified the foregoing’ refers to other content and requires it to be identified. ‘She confessed again’ may be the recorder's characterisation. Copy the formula into your record, explain its probable function and retain uncertainty when you cannot determine whether it summarises a more complex interaction.

Do not look only for traces that confirm pressure. Look also for the limits of your hypothesis. If a subject appears before any surviving question, that changes the argument about its introduction, without demonstrating absolute freedom. If an answer rejects a premise, it documents a relevant difference between question and reply. Critical reading improves when it allows resistance, negotiation, misunderstanding or initiative to be discovered, rather than reducing the entire document to machinery that always produces the same result.

The absence of questions may be a property of the documentary type. A record may be intended to preserve statements useful to the dossier, rather than reproduce the entire conversation. Compare documents of the same genre before treating an omission as exceptional. That comparison must remain contextual: similar formats in other records do not demonstrate that the sessions unfolded identically or that their writers followed exactly the same instructions.

5. Recognising voices without forcing word-by-word attribution

We can distinguish a voice attributed to the person making the statement, a voice organising the record and a voice interpreting it later. Each may contain material originating with other people. A statement includes rumours; a scribe incorporates the accused woman’s words; an editor preserves a procedural formula and adds a note. The text need not permit a perfect separation. Analysis aims to make recognisable layers visible and avoid attributions more precise than the evidence.

Begin with speech and knowledge verbs: said, heard, remembered, supposed, reported someone saying. These change testimonial scope. ‘Saw someone arrive’ and ‘heard that someone had arrived’ relate differently to events. Translating both as ‘knew’ loses information. Preserve attribution chains even when sentences lengthen. Stylistic economy must not erase distance between perception, rumour and deduction.

Then observe the categories. A word such as pact may appear as an authority’s classification, a formulation attributed to the person or an editorial title. All three occurrences are historically interesting, but they are not interchangeable. If the term appears in a modern heading, do not use it as automatic proof that the person described themselves with it. If it appears in a recorded answer, you must still examine interaction and transmission before presenting it as an entirely autonomous self-definition.

Revisable annotation

You can mark each fragment with simple working labels: attributed voice, recording formula, preserved intervention, editorial commentary, uncertain attribution. These labels belong to your workbook; they do not replace the quotation or constitute the protagonists' historical categories. They must be open to change. When a new comparison shows that a sentence was already in the manuscript, you will change its attribution without having to redo every observation about its content.

There is a risk of false precision: assigning percentages of authenticity to each sentence without a basis allowing them to be measured. Saying a passage is ‘seventy per cent popular voice’ creates a scientific appearance but explains no checkable operation. It is preferable to describe specific elements: attributed local vocabulary, a shared formulaic structure, an explicit question or an identified editorial intervention. Qualitative explanation can be rigorous when it allows another person to review exactly the same decisions.

Nor should individual voice be treated as a pure substance free from culture. People speak with learned words, shared narratives and social expectations. Finding a known motif does not demonstrate that the interrogator introduced it. Finding a distinctive detail does not demonstrate that it came from a literal experience. The relationship between cultural repertoire and personal initiative is a research question requiring finer arguments than an opposition between institutional copying and absolute spontaneity.

6. Pressure, voluntariness and diagnostic limits

The conditions of a statement cannot be reduced to the presence or absence of documented physical violence. Proceedings can include unequal authority, threatened consequences, dependence, repeated questions or expectations about what counts as an acceptable answer. These are analytical dimensions that we must investigate. Their general relevance does not demonstrate that each occurred in a particular form in Gowdie’s case. Always distinguish the repertoire of questions from the inventory of established facts.

The word voluntary must be read within its documentary use. It may characterise a statement according to procedural criteria without equating to our everyday idea of a conversation free from pressure. Before translating it as complete freedom, ask who writes it, what it is contrasted with and which facts the case file allows to be checked. A formula is not invalidated by being a formula, but neither does it replace a description of the session's material conditions.

In his study of language and beliefs in Scottish trials, Alaric Hall points precisely to the need not to infer the absence of all coercion from wording about the absence of torture. His analysis also shows how much an interpretation may change when a word's meaning is examined in context. We use it as an example of linguistic and documentary criticism, rather than as authorisation to assert that specific methods of pressure were used on Gowdie without having documented them. ^[3]

Avoid two shortcuts

The first shortcut is to declare a confession true because torture is not recorded. The conclusion introduces two leaps: from silence about a condition to its absence, and from that absence to the account’s accuracy. The second shortcut is to assert that everything stated is worthless because the relationship was unequal. Inequality requires us to study how the information was produced; it does not automatically eliminate its usefulness for every historical question.

The same sentence may have little value for demonstrating an extraordinary event and considerable value for studying an accusatory category. A list of names may not prove the actual participation of everyone mentioned, but can show whom the document includes within a network of suspicion. Preserve that distinction when presenting results. The source does not receive a single rating of truth or falsehood for every imaginable use.

Do not diagnose a historical person from the vividness of their narratives. Nor should the absence of a diagnosis become a demonstration of supernatural experience. Documentary work can describe narratives, categories and conditions without resolving clinical or metaphysical questions. If research proposes a psychological or anthropological interpretation, attribute it and examine the evidence it uses. Do not present the explanatory label as though it were transparently contained in the case file's words.

7. Repetition, ratification and corroboration

Four statements by one person are not four independent witnesses. They can provide different information, show changes and allow the development of the proceedings to be studied. But they may also depend on what was said earlier, on a reading of the previous record or on questions intended to complete it. The number of documents and the number of independent sources answer different questions. You must count both separately when independence affects your argument.

Repetition may mean a formulation persists. Ratification may document an institutional operation of confirmation. A new narrative may add elements absent earlier. None of these observations is sufficient to decide what happened outside the record. Compare precisely: are words, episodes, relationships or merely a general category repeated? Does the second session refer to the first? Do we preserve the same level of detail in both? Without these questions, a table of matches may exaggerate corroboration.

Imagine a modern summary copies a printed work, which reproduces two records in which the second ratifies the first. We have four reading objects, but a chain of dependence. Citing all four may help study transmission; it does not automatically multiply the evidence of the narrated event. Draw arrows between the documents and write the known relationship on each arrow: copy, summary, quotation, ratification or uncertain relationship. Doubtful arrows must remain visibly identified as such.

Changes needing explanation

A variant may result from clarification, omission, reformulation or editorial intervention. Begin by describing it before choosing a cause. If a new name appears in a session, write ‘the name enters the surviving record at this point’. Do not immediately write ‘the person finally remembered’ or ‘the interrogator imposed it’. Both explanations require additional information about the interaction, recording or transmission.

Nor are contradictions resolved by choosing the most detailed version. A detail may come from a later question, a genuine clarification or reorganisation. Compare each version's function and temporal relationship. Sometimes the correct conclusion is to preserve two incompatible formulations and indicate that the corpus does not allow a choice between them. A well-constructed case file may contain discrepancies; its quality depends on showing and explaining them as far as possible.

8. Reading the extraordinary as a historical problem

A narrative of transformations, encounters or actions impossible according to ordinary knowledge does not oblige historians to choose between literal credulity and contempt. We can investigate how it is articulated, what relationships it establishes, what vocabulary it uses and the situation in which it is recorded. These questions produce knowledge about culture and documentation without turning the account into a demonstration of powers or treating the person as a caricature. Respect does not require accepting every assertion as an event; it requires careful examination.

Narrative vividness has several components: specific scenes, names, sequences, sensory details and recognisable relationships. These devices may appear in memories, tales, rumours, doctrines or reconstructions under pressure. Vividness therefore does not function as independent evidence of literal truth. Nor does it allow us to infer that everything is invented. Describe what the narrative does: organises responsibilities, places participants, explains a misfortune or connects events. Then ask what external evidence would allow those relationships to be studied.

The Gowdie materials contain motifs that scholars relate to witchcraft and beliefs about fairy beings. The classification may be useful, but we must retain its analytical character. Two motifs placed under the same modern label did not necessarily have an identical meaning for every speaker. Comparison requires specifying what is compared: a word, a narrative sequence, a function or an interpretation. A general resemblance to another tradition does not, on its own, establish historical continuity or contact.

The temptation to reconstruct a ritual

A judicial narrative may mention actions and formulas. If we gather, order and complete them with other texts, we obtain a modern composition. That composition is not equivalent to the document and must be identified as such. Scholarly reading can describe what the record attributes without offering a contemporary recipe. Where order, conditions or context are missing, preserve the gaps. Completing a sequence to make it practicable modifies the object of study and may conceal its accusatory origin.

The same caution applies to the history of communities. A document’s list of participants does not demonstrate the existence of a stable organisation with the characteristics a present-day reader expects. It may record an attributed network, a judicial category or relationships of another kind. Studying a historical group would require documented connections, activities and context. Association within a confession is a starting point requiring verification, rather than membership we can retrospectively assign to every person named.

A useful conclusion may be modest and rich at once: ‘The document brings together certain motifs and articulates them within a procedural statement; their relationship to actual practices must be examined separately.’ This sentence does not empty the source of value. It opens several investigations: language, classification, social relationships, transmission and reception. Depth comes from distinguishing problems and constructing evidence for each, rather than replacing every question with a single spectacular explanation.

9. Fictional dossier: the Elm House statements

The following case is entirely invented. It neither reproduces Gowdie's trial nor seeks to reconstruct the conditions of a specific historical court. We have designed four documents to practise relationships of dependence. The fictional protagonist is called Inés. The matter concerns a meeting at Elm House. You do not need to decide whether that meeting happened; you need to establish what each document allows to be claimed and which changes appear as the information circulates.

Document A, first record. The heading indicates that Inés makes her statement before a local authority and a scribe. The body attributes to her having heard a conversation about a meeting in the house. It does not identify the person from whom she heard the news. A list of three names then appears, introduced by a formula indicating that she was asked who was there. The text ends with a statement of agreement attributed to Inés. No signature survives, nor any information about how that agreement was obtained.

Document B, second record, five days later. It states the earlier text was read to Inés and ratified, except she could not be certain of the third person’s presence. It adds a fourth name as someone mentioned in the conversation she says she heard. It describes no direct observation of the meeting. The closing passage preserves a formula of agreement similar to A’s. We do not have a complete transcription of the questions from either session.

Document C, a copy made several decades later. It reproduces A and B, but its copyist adds a marginal note: ‘participants in the meeting’. The note sits beside the four names. It neither explains that two names have different statuses nor preserves an unambiguous sign restricting the note to a single record. The copy declares that it comes from the earlier documents. Its value for studying later access is clear; its independence as a witness to the meeting is nil within the teaching case file.

Document D, modern publication. It summarises C by saying that Inés identified four attendees and confirmed their presence twice. The author does not cite A or B directly, but refers to the copy. The summary omits the distinction between having heard a conversation and having attended the meeting. It also omits the reservation concerning the third person. This publication is a new object for studying reception, but its assertion about the attendees depends on a reading that we can check against the rest of the dossier.

First solution

Draw up two lists. The first contains documentary facts: A attributes hearsay to Inés; B records a ratification with a qualification; C adds a note; D summarises more categorically. The second contains unproven external events: whether the meeting actually took place, each person's presence and the participants' intentions. The case file allows detailed analysis of the first list. It does not allow the second to be converted automatically into a factual narrative.

The following plate represents these relationships. Its letters identify documents in the exercise, rather than archival sources. Notice that information does not become more independent by moving to a new support. Its precision also changes: C's note groups names that A and B treated differently, and D removes qualifications. Locating the exact point of that change is more useful than describing the whole chain as true or false.

Visual reading of the fictional dossier

Four fictional documents are shown. A records hearsay. B ratifies with reservation and adds a name. C copies both and groups names in a note. D summarises C and loses qualifications. These are dependencies, not four independent meeting observations.

10. Solving the dossier step by step

The first decision is to choose a unit of analysis. If you investigate the third person's presence, find every statement referring to them and preserve its modality. A includes them in a list produced after a question. B introduces an explicit qualification. C groups names under a label. D asserts confirmed attendance. The sequence shows how an uncertainty disappears in transmission. It does not, on its own, demonstrate that person's actual situation outside the case file.

The second decision is to distinguish types of relationship. Between A and B, reading and partial ratification are declared. Between A/B and C, copying is declared. Between C and D, there is an editorial reference. These relationships have support within the fictional case. By contrast, the relationship between the conversation Inés heard and the meeting is an attribution whose intermediary we do not know. An arrow must name the operation; otherwise the diagram can suggest a chain of direct observations that never existed.

The third decision is not to confuse a qualification with a denial. ‘She could not be certain of their presence’ is not equivalent to ‘she was certain of their absence’. Turning the former into the latter replaces uncertainty with a negative claim. In a table, create distinct categories for attributed presence, attributed absence, uncertainty and unpreserved information. This choice affects every later summary and prevents readers from interpreting an empty cell as a demonstrated exclusion.

Defensible wording

An appropriate record would say: ‘The first record attributes to Inés information received in a conversation and records three names after a question about attendees. The second records a ratification with a qualification concerning the third name and adds another as a person mentioned. The later copy and its modern summary turn those unequal attributions into a uniform list. The case file provides no independent observation of the meeting.’ Each sentence can be connected to a document and makes clear what is not concluded.

A defective record would say: ‘Four documents confirm that four people attended.’ The error does not lie in the number alone. It confuses the existence of separate documents with their independence, hearsay with observed presence, qualification with confirmation and commentary with a statement. Adding ‘probably’ is not enough to correct it. The relationships removed by the sentence must be reconstructed. Precision depends on the argument's structure, rather than introducing a cautious word before a poorly formed conclusion.

What to seek next

The next search will depend on the question. To study the meeting, you would seek independent, relevant evidence about place, date and participants. To study the procedure, you would seek recording criteria and other comparable documents. To study reception, you would seek intermediate editions or quotations explaining the transformation. No expansion is universally better. Research advances when the new document responds to an identified uncertainty and its relationship to the rest can be described.

Even if a contemporary letter mentioning the meeting appeared, it would not automatically resolve everything. We would need to check whether it depends on the same rumour, whether it concerns the same gathering and what participation it attributes. Independence is not conferred by changing documentary genre. A letter, a record of proceedings and a chronicle may reproduce the same report. Analysis of their connections remains necessary even when their material forms differ and their authors are not the same.

11. Guided workshop: correcting excessive claims

First task. Review this invented sentence: ‘Because violence is not mentioned in the record, Inés spoke without pressure.’ Identify its two inferences and write an alternative. Before reading the answer, distinguish the document's silence from a positive description of conditions. Also ask what the case file would need to preserve for an absence to be informative. A record designed to register answers may not describe the session's circumstances in sufficient detail.

Reasoned answer. The sentence infers absence of violence from absence of mention, then turns that absence into complete freedom. A proportionate version would be: ‘The record consulted does not describe violence; that datum is insufficient to reconstruct all the conditions of the statement.’ The correction preserves what was observed and removes a conclusion the document does not support. Progress would require consulting specific information about the procedure, without automatically transferring conditions documented in other cases to Inés.

Second task. A student writes: ‘The appearance of the fourth name demonstrates that Inés remembered more details.’ Propose two alternative explanations compatible with the case file and an observation that would allow them to be distinguished. The aim is not to suspect everything, but to recognise that the same textual change may have several causes. An explanation becomes stronger when it provides evidence the alternatives do not explain equally well.

Reasoned answer. The name may have appeared after an unpreserved question or may have been a clarification offered by Inés. B may also have recorded information A summarised without including. A surviving explicit question or a draft available for comparison would help distinguish possibilities. With the current data, we can say only that B adds a name with a specific status: someone mentioned in conversation. Do not turn this into observed attendance or assign certain psychological causation.

Third task. C's marginal note is printed as a title in a new edition. What must a reading record point out? Explain why the change of position matters even though the words are the same. Then write a two-sentence editorial note preserving the information without attributing the heading to Inés.

Reasoned answer. The title may appear to belong to the original organisation and cover more content than the note. An appropriate note would say: ‘The heading reproduces an annotation in the later copy. It does not form part of the words attributed to Inés in the earlier records.’ The answer identifies origin and scope. If we could not establish the annotation's hand or date, that limitation would need to be stated rather than a precise authorship invented.

Fourth task. A reader proposes discarding the entire testimony because it does not preserve the complete questions. Explain which investigations it still allows and which remain limited. A sound answer mentions at least two positive uses and one limit. For example, we can study the surviving organisation of the accusation and the editorial transformation of qualifications; we cannot reconstruct a complete verbatim dialogue. The absence of one dimension does not destroy all the others.

12. Independent practice with explained answers

Exercises to complete before consulting solutions

Answer in your notebook. For each exercise, write a conclusion, the evidence supporting it and a limitation. Complete all the questions in this section before moving on to the answers, which begin on a new page. Keep your first answer so that you can compare it afterwards.

Exercise one. In a second fictional case, three publications cite the same paragraph from an old edition. Two use identical words; the third modernises the vocabulary. A student presents them as three corroborations. What minimal map should you construct? Before answering, separate publication history from the history of the narrated event. Modernisation may produce an appearance of independence that disappears when the references are reviewed.

Exercise two. A record attributes to a person having seen a light. A summary says they saw a supernatural figure. What changed? Answering ‘it was exaggerated’ is not enough. Identify the exact level of the change and explain which datum would support the new description. The difference may affect the perceived object, the observer's interpretation and the voice making the identification.

Exercise three. A formula says that the person confirmed ‘the foregoing’, but the copy has a gap before that sentence. Can you state what they confirmed? Explain how you would record the relationship and the distinction between a clear textual reference and materially lost content.

Exercise four. Two documents disagree about the number of participants. One is earlier and the other longer. Which do you choose? Answer without using a universal rule of age or length. Consider purpose, dependence, integrity and relationship to the narrated information.

Exercise five. Write a Gowdie sentence distinguishing record, publication and study.

13. Project: a confession with a reading dossier

Your final product will be an extended record of a single statement or a defined passage. Choose lawful access sufficient to identify the medium. It may be a course document or a public dossier suited to your reading level. You do not need to find a spectacular case. A short passage with questions, formulas and variants can offer more possibilities for analysis than a long narrative without context. Quality will depend on the precision and traceability of your decisions.

Begin with an identification: the person to whom the statement is attributed, the date of the record, the place where stated, the producing institution if known, and the edition or reproduction used. Distinguish these details from the dates of the narrated events. Add a note about access: image, transcription, translation or secondary quotation. This initial record prevents readers from attributing to your work contact with the original that you may not have had.

The project’s body will have four components. First, a brief paraphrase preserving attributions and modalities. Second, a table of voices and operations with located examples. Third, a map of dependence between the materials consulted. Fourth, a conclusion answering a specific question and identifying the missing information needed to go further. If you include a translation, explain which decisions affect the argument and separate your comments from the translated words.

Review criteria

Review argument-changing errors first. Incomplete references can be corrected; attributing an editor’s category to the person requires interpretative revision. Longer bibliography cannot compensate for attribution errors. Every reference needs a function and must support exactly its assigned point. If a source’s contribution is unclear, revisit the question before extending the list.

To close the project, add a short critical review: formulate the strongest objection a careful person might make. It may identify unresolved dependency, indirect access or an interpretative alternative. Respond with the available data and acknowledge what remains open. This exercise does not weaken your work. It distinguishes a robust conclusion from one that merely looks sound because it has never been tested against a specific alternative.

14. Glossary and writing decisions

Record of proceedings. A document that records proceedings through recognisable conventions. Its form may select and organise information. Before using it as verbatim dialogue, examine what it seeks to preserve and how it presents interventions, answers and closing formulas. The word does not, on its own, determine the level of detail or the fidelity of each passage.

Confession. A statement presented as acknowledging certain facts or responsibilities within a specific framework. The term may be a category of the procedure or an editorial characterisation. Its presence does not settle the historical truth of each part of its content or the conditions under which it was recorded. Identify who uses it and with what scope.

Ratification. An operation confirming previous content. To analyse it, you need to identify that content, the reservations accompanying the confirmation and the form in which it is documented. A ratification does not automatically constitute an independent observation. It can show continuity, modification or the closure of a stage in the proceedings.

Mediation. An intervention allowing information to be produced, preserved or transmitted: questioning, recording, copying, editing, translation or commentary. It is not equivalent to falsification. Studying it shows which transformations are possible and which can be checked. A mediated source remains a source; what changes is the type of access and the questions we can answer through it.

Modality. How sentences express certainty, doubt, possibility, obligation or attribution. ‘Saw’, ‘believes she saw’ and ‘heard tell’ are not interchangeable stylistic variants. Preserve them to retain testimonial status. Summaries particularly need verb and qualification review, since these often disappear before names or dates.

Independence. The absence of a dependency relevant to the claim being corroborated. It is not an absolute property conferred by having different authors. Two texts may be independent for some questions and share the same report for others. Independence must be argued in relation to a specific content and transmission relationship.

Documentary silence. Information absent from consulted material. Its strength depends on expected recording and access completeness. It may guide a search without establishing historical absence. Specify observation scope: fragment, edition, dossier or reviewed set.

Corroboration. Additional support for a claim through relevant evidence. It requires examining compatibility, dependency and scope. A verbal match may show copying rather than confirmation. An independent source may support only part of the claim. Explain which aspect has been strengthened and which remain unproven.

Sentences preserving complexity

When the document attributes an action, write ‘the record attributes the action to them’ if you have not established it by another route. When you encounter a formula, write ‘the record characterises the statement as’ and examine the meaning of that characterisation. When a study proposes an explanation, name the person proposing it. These formulas are not a ritual of caution: they correctly allocate responsibility for assertions.

When a question is missing, describe the transition and the hypothesis without inventing its words. When a variant adds information, name the change before explaining its cause. When you do not know an outcome, define the corpus and the search. When the account is extraordinary, retain its cultural interest and avoid turning narrative intensity into verification. The result will be more exact writing, also more useful to anyone wishing to continue the research.

15. Reference: breaking down and justifying claims

These tools adapted from lesson 5 review confession claims before publication. Apply them to “spoke freely and confirmed the facts”: separate conditions, content and corroboration. Manual and notebook examples are fictional methodological contrasts, not Gowdie evidence. They show one source supporting part of a sentence while leaving other parts unresolved.

Do not review a paragraph simply by asking whether it ‘covers the same topic’ as its bibliography. Two texts can discuss a city, a figure or a tradition without supporting the same claims. Checking begins when you turn the sentence into a question specific enough to answer with an identifiable passage.

Consider this fictional sentence: ‘The manual of 1860 proves that all initiates used this sign to protect their homes.’ It contains at least a document identification, a date, a population, a frequency, a behaviour and a purpose. An image of the sign in the manual verifies only one part. It does not automatically make the other five true.

Break down the statement: an identifiable manual exists; this copy dates to 1860; it contains the sign; it prescribes domestic use; its intended readers are initiates; all used it. The later claims need different evidence. Even if the manual commands use, whether anyone followed the instruction remains to be investigated.

A proposition is a claim that you can evaluate separately. You do not need to turn all your writing into lists of logical statements. Breaking a sentence down is a review tool: temporarily separate its parts to check them, then rewrite it clearly. This is particularly useful when words such as ‘always’, ‘all’, ‘original’, ‘proves’ or ‘continuity’ appear.

Three questions that are not equivalent

‘Does this source exist?’ checks the reference. ‘Does it contain that information?’ checks the attribution. ‘Does that information support my conclusion?’ checks the argument. Answering the first question does not resolve the second; answering both does not resolve the third. An impeccable bibliography can accompany flawed reasoning.

The scope of the statement also matters. ‘This copy contains a drawing’ refers to an object. ‘The edition contains a drawing’ refers to all its copies, although different states may exist. ‘The tradition used this drawing’ extends the scope to practices, places and people. Each extension creates a new burden of proof.

When reviewing, underline the verbs and circle the quantifiers. The verbs show which relationship you are asserting: mentioning, attributing, instructing, performing, transmitting, deriving. The quantifiers show how many cases you cover: one, some, most, all. Many exaggerations arise when you replace ‘prescribes’ with ‘proves that it was done’, or ‘one source’ with ‘the tradition’.

The table does not require a different document for each row. One source may support several propositions. What matters is that each has identifiable support and does not acquire the certainty of the others by association. If the date is secure and the use uncertain, that difference must survive in your writing.

Pause to check: rewrite the sentence about the manual, retaining only what a page showing the sign and its domestic instruction would establish. A sufficient answer would be: ‘The copy consulted prescribes the use of the sign to protect the home.’ Add the date only once you have checked its basis. Actual practice and its frequency remain open questions.

A conclusion needs a connection: the reason why a piece of evidence counts in its favour. If you find a recipe in a notebook, the evidence is its presence. To conclude that the owner carried it out, you must add an assumption: that they copied only actions they had performed. That assumption may be false. They may have been collecting readings, commissions or curiosities.

Making the connection explicit does not weaken the work. It makes it possible to check. Draft three sentences: ‘I observe this’, ‘I relate it to that for this reason’, ‘therefore, I support this conclusion within these limits’. Then turn them into natural prose. A clear argument matters more than repeating a formula.

In a fictional case, a notebook contains a recipe and a marginal note: ‘I tried it yesterday; the result was different.’ The note strengthens the evidence for an attempt, provided that its hand and date are relevant. It does not prove that the recipe worked. Nor does it prove that the whole community practised it. The connection must lead to the precise conclusion without extending beyond it.

Now consider another note: ‘They say it works.’ This sentence documents an attribution of efficacy and an indirect voice. It is not equivalent to a personal observation. A reader who removes ‘they say’ when summarising changes the kind of evidence. Attribution verbs, negations and qualifications are substantive information, not dispensable decoration.

The alternative explanation test

Ask what other situation would produce the same evidence. Two books share an image. There could have been direct copying, a common source, an intermediate reissue or a conventional composition. If all these possibilities fit, resemblance alone does not decide which occurred. The next step is to look for a feature that distinguishes between them.

This test does not require you to rule out every imaginable possibility. It requires attention to relevant alternatives compatible with the context and evidence. An explanation that needs many additional assumptions may be less convincing than another. But you must show what those assumptions are, rather than choosing the story that best fits your initial expectations.

You can also ask what would happen if the conclusion were false. If the same document could exist without the practice having been carried out, its existence does not establish performance. It might still provide context, vocabulary or a prescribed model. Identifying that specific value prevents you from discarding a source because it does not answer all your questions.

One warning sign is the phrase ‘this confirms what we already knew’ without identifying what it confirms. It may confirm the presence of a name, but not its interpretation. It may support a date of circulation, but not a date of composition. Replace ‘this’ with the evidence and ‘what’ with the specific proposition.

16. Reference: perception, memory and narration

This section adapts lesson 5 to preserve the differences between observation, recollection and received information. In a confession, pay particular attention to verbs connecting the person with what is narrated. The examples of letters and interviews allow practice outside the judicial case file. Afterwards, return to Inés and check what changed when her hearsay became a list of attendees.

A voice is not a transparent window

Testimony conveys information through a person in a particular situation. That person chooses words, remembers from a later point in time and addresses an audience. Recognising this mediation does not make the account useless. It allows you to ask what the person observed, what they interpreted and what they know through others. A single statement may contain all three.

Imagine a fictional letter: ‘Last night I saw a light in the drawing room. My sister said it was the sign our grandfather had foretold. Everyone was convinced.’ The first sentence reports a visual experience. The second conveys an interpretation attributed to the sister and a piece of family information. The third generalises about a collective mental state. Each proposition requires a different check.

To study the light, you would need information about the conditions of observation, duration and possible explanations. To study the family interpretation, the letter already offers a relevant attribution. To support the claim that everyone was convinced, you would need information about who was present and how they expressed their position. You should not grant one voice automatic access to everyone else’s mind.

Testimony may be sincere yet mistaken in an identification. It may also describe an observation correctly while offering an inadequately supported cause. Sincerity, perceptual accuracy and causal correctness are different dimensions. A discussion that asks only whether the witness is lying leaves much of the historical problem unexamined.

Temporal distance and changes in memory

The date of an account matters because a recollection is formulated from a later situation. A person may reorganise events in light of what they subsequently know. This does not mean that every late recollection is false. It means that you must distinguish the date of the event, the date of the earliest surviving report and the date of the version you consult.

Fictional case: an autobiography of 1910 describes a meeting in 1870. A letter of 1871 alludes to that meeting with fewer details. The autobiography adds a dialogue and a name absent from the letter. Their absence does not prove that the details were invented, but the letter’s proximity in time does not automatically confirm the fuller version either.

Compare their functions. The letter may have been written to report on a journey, while the autobiography constructs a personal conversion story. The difference in genre may explain selection and emphasis. Then investigate which details have additional support. Do not turn the comparison into a contest in which one document must be wholly true and the other wholly false.

The earliest surviving version is useful but need not be the first that existed. “The earliest report located in this dossier” bounds the finding to completed work. “The story began in 1910” would require proving no earlier circulation, a much harder claim.

Questions, interviews and prompted answers

An answer may depend on how the question was phrased. ‘What happened?’ opens a different field from ‘When did the figure appear?’ The second presupposes an appearance. If only the answer survives, you may lose part of the context that produced it. In an interview case file, record the questions and conditions when available.

You do not need to attribute manipulation to every interviewer. Questions organise any conversation, even in good faith. The task is to recognise which categories the questioner supplied and which the interviewee used spontaneously. This distinction is particularly useful when studying experiences that people interpret through different vocabularies.

Fictional example: three interviews use the word ‘initiation’, but in each it first appears in the question. This agreement does not, on its own, establish that all three people named their experience that way before the interview. It may document acceptance of the term, adjustment to the exchange or an already established use. You must examine the context to decide what you can support.

For interviews, identify recording, revised transcript, summary or published extract. Each preserves different information. Summaries may omit hesitation or reformulation. Interview editing mediates just as manuscript editing does.

Testimonies that depend on one another

Two people may discuss the same event after talking to each other. Their accounts remain distinct documents, but some agreements may derive from that conversation. Independence requires studying relationships, not counting names. Later agreement may be evidence of shared memory rather than two separate observations of the same detail.

In a fictional workshop, Elena writes a note on the night of a meeting. Weeks later, Daniel writes his recollection after reading it. Both mention an identical phrase. To investigate what was said at the meeting, you must consider this dependence. To study how the group consolidated a version, the note’s circulation is central evidence.

Nor does a discrepancy always amount to a decisive error. Two people in different positions may observe different aspects. One remembers the beginning; another arrives late. Comparison should seek real incompatibilities on the same point, rather than penalising every difference. You cannot demand confirmation from a source of something it was not in a position to observe.

Conclusions appropriate to the testimony

A sound formulation might say: ‘The letter reports its author’s observation of a light and records her sister’s interpretation. The conclusion about the conviction of everyone present belongs to the narrator and is not supported here by individual testimonies.’ The sentence preserves a narrated experience, an attribution and a limit without ruling on matters that the case file cannot resolve.

To study beliefs, you can analyse how the interpretation is used, which vocabulary it draws upon and which social relationships it expresses. You do not need to establish a supernatural cause to recognise the account’s historical interest. Nor should you present the narrator’s explanation as an external fact. Describing an attribution precisely is itself a scholarly result.

17. Reference: silences and information states

Tools from lessons 5 and 17 are brought together here to read omissions and record uncertainties. An unpreserved question, an illegible answer and an explicit denial require different fields. Apply the distinction to the conditions of statements and judicial outcomes. Do not complete a field from what usually happens in other cases; preserve the distinction between general context and individual information.

‘It does not appear’ needs a qualification: where does it not appear, in what form, and after what search? Failing to find a name on one page is not the same as failing to find it in a complete work. Nor does failing to find it through an automated search establish its absence from a manuscript with variant spellings.

Begin by distinguishing three situations. In the first, you have reviewed a complete document and the information is absent. In the second, you have only a fragment. In the third, the text exists but your access is inadequate: pages are missing, the image is illegible or the search does not recognise the script. Each situation permits different wording.

The strength of an absence depends on the expectation of presence. An inventory that claims to record every book in a room may provide relevant information about a missing title. A letter devoted to another subject offers a much less informative absence. Before arguing from silence, explain why you would expect to find the information there.

Consider two fictional exercises. A complete attendance list for a meeting does not include Marina. A report about the decoration at the same meeting does not mention her either. The two absences do not have the same value. The list aims to enumerate participants; the report might omit them all. Genre and purpose alter the expectation.

Even the list requires checking. Is it complete? Does it cover unregistered visitors? Was it written before or after the event? Could Marina appear under another name? The aim is not to invent excuses for every absence. It is to identify the conditions that allow absence to become evidence and recognise when those conditions are not met.

In a fictional exercise, a particular dedication does not function as an inventory of every figure worshipped throughout every period of a city's history. ‘That name does not appear here’ does not imply ‘that cult never existed there’. The universal conclusion exceeds the observation's temporal and documentary scope. The same applies when a confession omits a name: first examine what it sought to record and which part of the case file survives.

How to record an unsuccessful search

Note the body of material examined, its limits and your procedure. For example, in a fictional case file: ‘Pages 1 to 32 of the digitised copy were reviewed; pages 33 and 34 are missing. The term was not located in the available portion.’ This sentence allows another person to extend the search without having to reconstruct what you did.

If you work with automatically recognised text, search for variants and check the relevant images. Misread letters, line-end hyphens and historical spellings produce false negatives. A computer search result first describes a search operation; turning it into a documentary claim requires examining its limitations.

An absence can contribute to a conclusion alongside other evidence. Reasoning from silence is not forbidden. You need to justify the expectation, define the body of material and consider preservation problems. A conclusion becomes stronger when several relevant checks converge, not when you repeat ‘it does not appear’ more emphatically.

An empty cell may mean unknown information, a datum not yet entered or a question inapplicable to the record. Do not choose its meaning by intuition. Before analysing a table, consult the conventions its manager uses. In your own case file, expressly distinguish missing information, documented denial and an inapplicable field; that difference protects interpretation of the results.

This convention changes the calculations. If an occupation field is empty, you should not classify the person as someone with no occupation. If an indicator is not marked, you cannot automatically conclude that the event was absent. You must read the field’s definition and the type of record. A box’s binary appearance may conceal an epistemological distinction: we know that evidence exists, or we do not have it. It is not necessarily an opposition between an event that happened and one that did not.

Four conceptual states

Use fictional residence data. ‘Known residence’ contains a place. ‘Unknown residence’ lacks information. ‘Not applicable’ means the record does not represent a person for whom residence makes sense. ‘Documented denial’ requires explicit rejection of a proposed residence, such as stating the person did not live there. Do not merge these states for spreadsheet convenience.

This table is a model of reasoning, rather than a claim that every project uses exactly those four labels. Each project's documentation determines how they are represented. If you export data and replace blanks with a word, preserve the original meaning. Do not silently turn every state into zero to facilitate operations. That convenience may change the historical question before the first calculation is made.

Error and repair example

An invented set has one hundred records, only thirty with identified occupation. Twelve mention textile work. Twelve divided by thirty describes the proportion among known occupations; twelve of one hundred describes records containing that mention. You cannot say the other eighty-eight people did not work in textiles. Seventy have unknown occupations; eighteen have other classifications. These are different groups.

The correction consists in showing three quantities: mentions of textile work, other known occupations and unknown occupation. Then define the denominator of the percentage you use. A chart hiding the third group may suggest complete coverage. A clear note must explain that the distribution describes the subset with information. Missing data are not a nuisance removed at the end; they form part of interpreting the result.

When to pause comparison

If two regions have very different proportions of known information, direct comparison of categories may mislead. Before explaining social differences, compare documentary coverage. Sources may record occupations in one region and barely do so in another. This possibility does not demonstrate that every difference is documentary, but requires examination. A responsible conclusion distinguishes the observed pattern from the conditions that may have produced it.

18. Reference: people, sessions and documents

This adaptation of lesson 17 prevents the number of documentary pieces from becoming the number of people or confirmations. For the historical case's four statements, prepare separate identifiers for person, session and editorial witness. The quantities answer different questions. The following workshop uses invented examples to show how a single biography may generate numerous rows without multiplying historically.

A person is a biographical entity; a case organises a history of accusation; proceedings describe a course of action; a document preserves information about some part of that history. These units are related, but cannot be counted as equivalents. A person may be linked to several actions, and each action may generate more than one document. Joining those tables without controlling the relationships may turn one biography into many rows.

In your own case file, define session, action and proceedings separately. Do not record an action solely because it seems likely to have taken place. Explain how you group documents under a case and which evidence allows them to be related. The model must not become an automatic translation of everyday vocabulary. Its terms need definitions another person can apply to the same materials.

A complete example

Imagine a fictional person, Elspeth A. A case file preserves a denunciation, an authorisation, two judicial actions and a later note. There is one person, one teaching case, two actions and five documentary pieces. If a table has one row per piece, counting five rows as five people would multiply the population by five. If an export repeats biographical data beside each action, counting rows as cases would also produce an error.

The solution is not to delete repeated rows automatically. They may represent different events and preserve necessary information. Choose the level of analysis. To count people, use a person identifier and check its uniqueness. To study actions, preserve their own identifiers. To evaluate documentation, count pieces using explicit criteria. The same database may produce several correct totals because it answers different questions. The error arises when the result's name does not match the unit counted.

Identifier and name

An identifier allows a record to be referred to without relying on its visible name. Two people may share a name, and one person may appear under different variants. Grouping by the text string displayed on screen is therefore insufficient. Nor does the identifier by itself guarantee that the historical identification is perfect. It represents a project decision that can be revised if new evidence emerges. You must retain both its technical usefulness and the possibility of investigating the identification.

Fictional example: a person makes statements in two sessions of the same investigation. The continuity of the dossier allows them to be grouped, but each session retains its date and content. More than one procedural action does not require the creation of more people or the multiplication of the narrated events. Before counting, you must explain the relationship you are representing and the evidence supporting it.

Check before any percentage

Write down the total of distinct identifiers at each relevant level. Then explain why the quantities may differ. If you cannot, review the export or query. This check prevents errors a chart would later make difficult to detect. A visually convincing result may conceal a multiplication of rows. Traceability begins at record level, long before colours or titles are chosen.

19. Reference: preservation and representativeness

Lesson 17 provides a check on scale here. An extensive confession allows many details to be studied, but its richness does not make it a representative sample of all accused people. Before comparing motifs or outcomes, review which documents could have survived and which fields each project recorded. The numerical examples are fictional and serve to distinguish documentary coverage from historical frequency.

Missing information is not always distributed randomly. One type of court may leave more documentation than another. One period may be better preserved. An exceptional case may produce more writing and attract more editorial attention. These possibilities affect the relationship between known records and the historical population. Acknowledging missing data is not enough: ask which conditions may have favoured the survival of some rather than others.

The distinction between quantity and representativeness is decisive. A large subset may poorly represent certain dimensions if its selection is related to them. Conversely, a small group may offer valuable information about a specific procedure without representing the whole persecution. The question must fit the coverage. Do not use ‘many cases’ as a guarantee that any percentage describes the historical universe. Size does not automatically correct uneven selection.

Controlled comparison

Consider two fictional regions with one hundred recorded people in each. In region A, twenty outcomes are known, ten of them executions. In region B, eighty outcomes are known, forty of them executions. Among known outcomes, both show fifty per cent. Across all records, documented executions represent ten per cent in A and forty per cent in B. The difference between those last figures may reflect coverage, rather than an equivalent difference in historical outcomes.

Nor can you conclude that the regions had exactly the same actual proportion. You have only shown that the available information permits a documentary explanation for part of the contrast. To proceed, you need to study how outcomes were preserved and what types of cases each subset contains. The exercise teaches a precise caution: a difference between records does not automatically become a difference in the phenomenon. Nor is it automatically eliminated by suspecting bias.

Describing missing data

Prepare a coverage table by period, region or source type, according to your question. State how many records have information and how many lack it. Then compare the proportions before studying the main characteristic. This table may be less striking than the results chart, but explains its conditions. If a category appears only in detailed case files, acknowledge that your search may be selecting documentary richness as well as the subject under investigation.

It is also useful to distinguish lack of knowledge about the event from lack of knowledge about the coding. A field may be empty because the source does not contain the information or because the project did not collect it systematically. Do not attribute to the archive what results from a design decision. Review the recording instructions and the fields the team used unevenly. The existence of a column does not guarantee that it can support a quantitative comparison.

When you cannot correct it

You will not always be able to estimate how much information is missing or reconstruct its distribution. In that case, limit the conclusion to the observed corpus and explain the problem. You can formulate preservation hypotheses and identify documents that would help evaluate them. Do not invent correction weights to produce an apparently definitive figure. An adjustment needs assumptions and evidence. Applying it without a basis may introduce more error than it seeks to remove.

Useful wording would be: ‘The distribution describes records with available information; coverage varies between regions and does not permit direct extrapolation.’ This sentence must be accompanied by the corresponding quantities. The limit then functions as a specific property of the result, rather than a generic warning appended at the end. Readers know what they can use and which comparison needs further work.

20. Reference: reading studies while keeping the person in view

This closing section adapts lessons 5 and 17. Use it when reading a specialist interpretation of confessions: identify question, corpus, procedure and conclusion, then check how it represents the people. An influential explanation still needs attribution and evidence. Your synthesis may present a controversy without pretending to have resolved a linguistic or documentary question for which your access is insufficient.

A scholarly article is not a box of facts ready to copy. It is a contribution that selects a problem, presents materials and argues for an answer. Using it well requires recognising that structure. Reading only the abstract may identify the topic, but rarely suffices to evaluate a specialist proposal. The abstract announces; the main discussion shows why the author considers the conclusion reasonable.

Start by separating four components: question, corpus, procedure and result. The question defines what the researcher wants to find out. The corpus identifies the documents examined. The procedure explains how they are compared or interpreted. The result states what the author believes has been established. Mixing these components can lead you to cite an open question as an answer or turn a working hypothesis into a finding.

Reconstructing the argument’s structure

Imagine a fictional article studying five copies of a manual and proposing that two belong to the same branch. Its question concerns transmission relationships; its corpus is those five copies; its procedure compares variants; its result is a reasoned grouping. It has not necessarily studied every reader of the manual or all their practices. A valid conclusion about copies does not automatically extend to communities.

Write the argument in three lines: ‘The author observes…’, ‘considers that this observation indicates…’, ‘and therefore proposes…’. This small outline often reveals where you need to read more carefully. If you cannot complete the second line, you have not yet understood the connection. Copying only the result would leave you unable to explain why it is convincing or open to challenge.

Then distinguish what the article takes from other studies and what it contributes. A date may come from an earlier catalogue; a transcription may reuse an edition; a comparison may be original. Citing the article you consulted is correct when describing its argument, but you should not present it as an independent physical examination if it merely relays information from elsewhere.

What to do when two specialists disagree

Before choosing a position, check whether they really answer the same question. One author may study a name’s meaning in a particular passage and another its history over several centuries. Their different answers do not necessarily constitute a contradiction. They may also use different corpora or define terms such as ‘origin’, ‘use’ or ‘tradition’ differently.

If the disagreement is real, locate the point of divergence. Do they read a letter differently? Accept different dates? Give different weight to a resemblance? Use incompatible explanatory models? Locating this point enables an informative comparison: not ‘there are opinions to suit every taste’, but ‘the difference depends on how this feature is identified and which parallels are considered relevant’.

Do not settle a technical controversy by counting authors on each side if you do not know how their arguments relate. Several publications may repeat a proposal without new evidence. Nor should you claim consensus because an introductory manual presents an answer without discussion. Assessing consensus requires examining how the issue is treated in relevant, up-to-date research.

When expertise does not permit assessing specialist readings, useful synthesis remains possible: attribute proposals, identify published reasons and specify what further examination is needed. Honesty distinguishes presenting controversy from resolving it; it does not require silence. “I cannot adjudicate the palaeography” is precise when followed by the actual palaeographical issue.

A reading note that remains useful

Record the full title, version, pages used and a sentence of your own about the argument. Add two mental columns: ‘what this study allows me to cite’ and ‘what I have not checked’. The first may contain a specific proposal and its reasoning. The second may contain the contents of mentioned works that you have not opened. This prevents a long bibliography from creating a false memory of reading you have done.

End the note with a decision: use, compare, investigate further or set aside for this question. ‘Set aside’ does not mean the article is poor; it may not address your problem. Efficient research does not collect references for their quantity. It builds a set in which each item has an identifiable function and can be retrieved when it is time to write.

Abstraction allows groups to be studied that individual reading could not encompass. But each row refers to a history whose documentation is uneven and whose interpretation requires care. Preserving the case does not mean abandoning counting. It means keeping units, mediations and routes back to the sources visible. A number gains value when we know what it represents and what it has left out. This precision also avoids reproducing the documents' accusations as identities.

Review at three scales

At record level, check identity, fields and references. At group level, review coverage, relationships and denominators. At explanatory level, assess which mechanism and context support the interpretation. Work may be correct at one scale and fail at another. An accurate record does not guarantee appropriate aggregation; a correct sum does not guarantee a cause. The review must cover all three.

You can use this route whenever you encounter a historical visualisation on a website, in an exhibition or in a class. First identify what is counted. Then ask which records do not appear. Finally, examine the leap connecting the pattern to the narrative. You do not need immediate access to the entire archive to detect an excessive claim. You do need to distinguish a verified objection from a pending question. Criticism, too, must fit its evidence.

What a well-explained data library offers

A useful entry preserves definitions, examples, results and limits accessibly. It allows one reader to locate a number quickly and another to examine its construction in depth. It does not force a choice between a superficial summary and a case file impossible to consult. Editorial design can organise reading levels: initial orientation, development, workshop and references. Depth becomes usable when important decisions are explained where they affect interpretation.

The same care applies to a material measurement and a statement. In both cases there is a chain between evidence and conclusion. A composition may support the origin of a raw material without revealing who owned the object; a proportion may describe a corpus without reconstructing the entire historical population. The shared competence consists in stopping each claim at the limit of the available evidence.

Retain one concrete practice: never publish a historical figure unless you can explain its unit, base and provenance. Then add the question that the figure helps to investigate. The number thereby ceases to function as an authoritative way of closing the conversation and becomes a shared tool. History remains present because readers can return from the summary to the record and from the record to the document that supports it.

21. Sources and research routes

[1] Pitcairn, Robert, ed. Ancient Criminal Trials in Scotland. Edinburgh, 1833. Historical reference in the editorial corpus. The Museum of Witchcraft and Magic record “Isobel Goudie’s (Gowdie) 1662 Confession from Pitcairn’s Criminal Trials in Scotland” was checked. It establishes an editorial route; no complete independent print–manuscript collation is presented here.

https://museumofwitchcraftandmagic.co.uk/library/1390-isobel-goudies-gowdie-1662-confession-from-pitcairns-criminal-trials-in-scotland/

[2] Wilby, Emma. The Visions of Isobel Gowdie: Magic, Witchcraft and Dark Shamanism in Seventeenth-Century Scotland. First edition: Sussex Academic Press, 2010, according to the source editorial corpus. JSTOR electronic version consulted, published by Liverpool University Press, copyright 2013. DOI 10.2307/jj.3485506. The publisher’s presentation, contents and publicly accessible extracts were consulted, not the complete text. Authentication in progress is mentioned in that presentation, not asserted here as current status.

https://www.jstor.org/stable/jj.3485506

[3] Hall, Alaric. “Getting Shot of Elves: Healing, Witchcraft and Fairies in the Scottish Witchcraft Trials”. Folklore 116, no. 1, 2005, pp. 19–36. University of Glasgow repository version. Its analysis of language, categories and interpretative conditions was consulted. Read arguments with examples and references; they authorise neither retrospective diagnosis nor complete reconstruction of every interrogation.

https://eprints.gla.ac.uk/3081/1/getting_shot1.pdf

Extending the research

Start by locating the document and edition used. To study a particular reading, consult the witness or an apparatus that allows it to be checked. To evaluate an interpretation by Wilby, read the relevant chapter, rather than only the publisher’s presentation. To discuss the conditions of a session, look for specific documentation and distinguish absence of mention from positive evidence. Outcomes require their own documents; they cannot be inferred from the intensity of the confession. The development retains the core of the editorial DOCX for The Method Files lessons 23 and 24. The reference notebook adapts materials from lessons 5 and 17, with provenance indicated in each section. Elm House, Inés and their documents are fictional. They do not constitute new information about Gowdie. Digital references were checked in October 2026.

Sources for this reading

1. Pitcairn, Robert, ed. Ancient Criminal Trials in Scotland. Edinburgh, 1833. Historical reference in the editorial corpus. The Museum of Witchcraft and Magic record “Isobel Goudie’s (Gowdie) 1662 Confession from Pitcairn’s Criminal Trials in Scotland” was checked. It establishes an editorial route; no complete independent print–manuscript collation is presented here. Consult the source
2. Wilby, Emma. The Visions of Isobel Gowdie: Magic, Witchcraft and Dark Shamanism in Seventeenth-Century Scotland. First edition: Sussex Academic Press, 2010, according to the source editorial corpus. JSTOR electronic version consulted, published by Liverpool University Press, copyright 2013. DOI 10.2307/jj.3485506. The publisher’s presentation, contents and publicly accessible extracts were consulted, not the complete text. Authentication in progress is mentioned in that presentation, not asserted here as current status. Consult the source
3. Hall, Alaric. “Getting Shot of Elves: Healing, Witchcraft and Fairies in the Scottish Witchcraft Trials”. Folklore 116, no. 1, 2005, pp. 19–36. University of Glasgow repository version. Its analysis of language, categories and interpretative conditions was consulted. Read arguments with examples and references; they authorise neither retrospective diagnosis nor complete reconstruction of every interrogation. Consult the source