Skip to content
Back to search results

Archival Content Analysis

Introduced by
Adapted from Berelson and Lasswell content analysis; Furet's quantitative history
Year
1952
Type
mixed qualitative-quantitative method
Family
content analysis
Font size

What does it do?

Archival content analysis adapts the social-scientific technique of content analysis to the systematic study of historical documents held in archives.

Where the impressionistic reading of sources risks privileging the vivid or the convenient, content analysis imposes an explicit, replicable procedure: a defined corpus, a coding scheme of categories, the consistent application of those categories to every document, and the analysis of the resulting frequencies and co-occurrences. Pioneered for mass communication by Bernard Berelson and Harold Lasswell, the approach was absorbed into the quantitative history championed by Francois Furet and others, who treated runs of administrative records as data to be counted and tabulated. Applied to archives, however, the method must reckon with a complication absent from designed surveys: the archive was not created to answer the historian's questions. Its categories, survivals, and silences reflect the purposes and power of the institution that produced it, so disciplined coding must be paired with critical reflection on the archive's own logic.

The intuition behind the method

Suppose you want to know how attitudes to poverty changed over a century, and you have ten thousand parish relief records. Reading a memorable few and generalizing would be unreliable; the striking cases may be atypical, and your impressions unverifiable. Content analysis offers a more honest path. You decide in advance what to look for, perhaps the stated reasons for granting or refusing relief, define clear categories for each, and then apply those categories uniformly to a representative sample of records. Now you can count: how often each reason appears, how the mix shifts over time, which reasons cluster with which applicants. The discipline lies in making the rules explicit before you start and in checking that another reader, applying the same rules, would code the documents the same way. Yet a further subtlety distinguishes archival from ordinary content analysis. The records exist because an institution chose to keep them in a particular form for its own ends. The categories you find, and the people who never appear, are artifacts of that institutional gaze, and your analysis must read both the content and the archive that shaped it.

History

Content analysis took shape in the early twentieth century with the quantitative study of newspapers and was codified during and after the Second World War by Harold Lasswell's propaganda analysis and Bernard Berelson's Content Analysis in Communication Research (1952). Klaus Krippendorff later gave it a rigorous methodological foundation. In history the technique converged with the quantitative turn of the 1960s and 1970s, as Francois Furet and the Annales school promoted the counting of serial documents and the treatment of archives as data. The later cultural and archival turns, influenced by Michel Foucault and by Ann Laura Stoler's call to read along and against the archival grain, tempered the positivism of early content analysis with critical attention to the archive as an instrument of power and classification.

When to use it

Use archival content analysis when a large or repetitive body of documents could answer a research question that impressionistic reading cannot reliably settle, and when the documents are regular enough to be coded into consistent categories. It suits administrative records, registers, petitions, trial records, newspapers, correspondence series, and any corpus where frequencies and trends matter. The method is valuable for tracking change over time, comparing groups or places, and providing a systematic evidentiary base for arguments about attitudes, practices, or social composition. It complements serial history, which counts homogeneous facts, and prosopography, which codes biographical attributes. It is less apt for tiny corpora, where close reading suffices, or for highly heterogeneous sources resistant to a common scheme.

Limitations and considerations

  • Coding can strip statements from the context that gives them meaning, flattening nuance
  • Results are only as representative as the surviving, preserved sources, which are rarely a neutral sample
  • Constructing and validating a reliable scheme and coding a large corpus is highly labor-intensive
  • Frequencies measure what was recorded, not necessarily what occurred, conflating reality with the archive

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Replaces selective impression with a systematic, replicable procedure over an entire corpus
  • Reveals frequencies, trends, and co-occurrences invisible to unaided reading of individual documents
  • Intercoder reliability checks make the coding transparent and reproducible by other scholars
  • Bridges qualitative meaning and quantitative pattern, suiting mixed-methods historical research

Applications

  • Tracking shifts in the stated grounds for poor relief or criminal sentencing across decades
  • Measuring thematic and tonal change in newspapers, pamphlets, or political speeches over time
  • Coding petitions or correspondence to map grievances, networks, and rhetorical strategies
  • Analyzing inquisitorial or court records for recurring charges, defenses, and social patterns
  • Building structured datasets from registers to feed serial-historical and prosopographical analysis

How it works

How it works

1

The analyst first defines the corpus and the sampling frame: which documents, from which series and dates, will be analyzed, and on what principle a manageable subset is selected if the whole is too large. This requires understanding the archive's structure, how the records were generated, classified, and preserved, so that the corpus is coherent and its representativeness can be assessed. Decisions about inclusion, exclusion, and sampling are documented, because they bound every conclusion the analysis can support.

Formula
\mathcal{D} = \{ d_1, \ldots, d_N \}, \quad d_i \sim \mathrm{archive}, \ \mathrm{sampling\ frame}
2

Next the analyst constructs a coding scheme, a set of categories that operationalize the research question. Categories should be clearly defined, ideally mutually exclusive and collectively exhaustive within each dimension, and grounded in both the research aims and a pilot reading of the sources. Schemes may be deductive, derived from theory, or inductive, emerging from the documents, or a blend. A codebook records each category's definition, decision rules, and examples, so that coding is consistent and the scheme is transparent to other scholars.

Formula
\mathcal{C} = \{ c_1, \ldots, c_K \}, \quad \mathrm{mutually\ exclusive}, \ \mathrm{exhaustive}
3

The analyst then applies the scheme to every document in the corpus, recording for each whether a category is present or counting its occurrences, and capturing relevant attributes. This is the labor-intensive heart of the method, demanding careful, consistent reading. Coding may be manual or computer-assisted, and the resulting matrix of documents by categories becomes the dataset. Throughout, the coder must respect each document's meaning in context, since mechanical pattern-matching divorced from interpretation produces misleading counts.

Formula
X_{ik} = \mathbb{1}[\,c_k \in d_i\,] \ \text{or}\ \mathrm{count}(c_k, d_i)
4

To guard against idiosyncratic reading, the analyst assesses coding reliability, typically by having two or more coders independently code a subset and measuring their agreement with a statistic such as Cohen's or Krippendorff's kappa that corrects for chance. Low agreement signals ambiguous categories or unclear rules, prompting revision of the codebook and recoding. Establishing reliability makes the coding reproducible rather than the private impression of one reader, which is the methodological gain that distinguishes content analysis from ordinary close reading.

Formula
\kappa = \frac{p_o - p_e}{1 - p_e}, \quad \text{intercoder agreement}
5

Finally the analyst examines the coded data for patterns: the frequencies of categories, their change over time, and their co-occurrence and association. Quantitative summaries and, where warranted, statistical tests reveal trends and structures invisible to unaided reading. Crucially, these patterns are interpreted in light of the archive's own logic, asking how the institution's categories and selective preservation shaped what could be counted. The numbers are thus read as traces of both past reality and the recording apparatus, not as transparent measurements.

Formula
f_k = \tfrac{1}{N}\sum_i X_{ik}, \quad \mathrm{patterns} = g(X) \mid \mathrm{archival\ logic}

Common pitfalls

  • Mistaking the archive's categories and survivals for an unbiased picture of the past
  • Counting occurrences without interpreting them, producing precise but meaningless tallies
  • Designing categories that are vague or overlapping, yielding low and unstable intercoder agreement
  • Generalizing from an unrepresentative sample shaped by selective preservation and access

Frequently asked questions

How does archival content analysis differ from ordinary close reading?

Close reading interprets selected passages in depth but cannot guarantee that its examples are typical. Content analysis applies an explicit coding scheme uniformly to a defined corpus, allowing frequencies and trends to be measured and replicated. It trades some interpretive richness for systematic coverage and reproducibility. The two are complementary: close reading often informs the coding scheme, and content analysis tells you whether close-read examples represent broader patterns.

What does it mean to read the archive's own logic?

It means recognizing that archives were created by institutions for their own purposes, not to answer historians' questions. The categories the records use, what they record about whom, and which documents survive all reflect the institution's gaze and power. Coding the content alone risks treating these artifacts as neutral data. Reading the archive's logic asks who created the record, why, and who is absent, so that patterns are interpreted as traces of both events and the recording apparatus.

Is intercoder reliability necessary if one researcher does all the coding?

It is still valuable. Even a single coder can drift over time or apply ambiguous rules inconsistently. Having a second coder independently code a subset and computing an agreement statistic tests whether the categories are clear and the coding reproducible. High agreement supports the credibility of the results; low agreement reveals that the scheme needs sharpening. Reliability checking is what separates systematic content analysis from idiosyncratic impression.

Related methods

Sources

  1. Furet, F. (1971). Le quantitatif en histoire. In J. Le Goff & P. Nora (Eds.), Faire de l'histoire (Vol. 1, pp. 42-61). Gallimard.

    ISBN: 9782070287666
    Copied.
  2. Howell, M., & Prevenier, W. (2001). From Reliable Sources: An Introduction to Historical Methods. Cornell University Press.

    ISBN: 9780801485602
    Copied.