All subjects · English Language · WEN04 Investigating Language

Quantitative analysis: counting fairly

Edexcel IAL English Language · WEN04 Investigating Language

Numbers make a Section B argument persuasive: 'my data showed' is stronger than 'I think'. But counts must be fair. Two rules make them fair: count against every possible occurrence, and normalise for text length.

Practise this in the app: instant marking and a fix for every mistake. Free, no email needed to start.

Practise this lesson

The Principle of Accountability

Essentials of Linguistics explains that when you count a variant (such as -ing said as [ɪn]) you must count it against every occurrence of the variable (all -ing words, whichever way they were said). Raw counts mislead. Their example: a speaker uses [ɪn] 10 times in a reading passage and 8 times in casual speech, which looks like more in reading; but out of all -ing tokens it is 25% in reading and 40% in casual speech.

Normalised frequency

Texts differ in length. A Mainz University corpus page explains: divide the raw count by the total words and multiply by a common base (per 1,000 or per million words). Their example: 'awesome' occurs 1,133 times in about 95.6 million words of spoken American English, which is about 11.86 per million words.

Useful measures for Unit 4

Always say how much data you had: a pattern in two texts is a hint, not proof.

Worked example

Leader A's email has 18 uses of 'we' in 1200 words. Leader B's chat posts have 9 in 300 words. Who uses 'we' more?

  1. A: 18 / 1200 x 1000 = 15 per 1,000 words.
  2. B: 9 / 300 x 1000 = 30 per 1,000 words.
  3. Raw counts suggested A; normalised counts show B.

Answer: Leader B uses inclusive 'we' about twice as often (30 vs 15 per 1,000 words), despite the lower raw count.

Key idea: Count variants against all occurrences (accountability). Normalise by length (per 1,000 or per million words). Report the size of your data and treat small samples cautiously.

Check yourself

  1. A speaker says [ɪn] 10 times out of 40 -ing words in a reading passage. What percentage is that?
    Show answer

    25 %. 10 / 40 x 100 = 25%.

  2. In casual speech the speaker says [ɪn] 8 times out of 20. What percentage?
    Show answer

    40 %. 8 / 20 x 100 = 40%.

  3. 'Awesome' occurs 1,133 times in 95,565,075 words. What is the frequency per million words (2 dp)?
    Show answer

    11.86. 1133 / 95565075 x 1,000,000 = 11.86.

  4. A chat log has 9 uses of 'we' in 300 words. How many per 1,000 words?
    Show answer

    30 per 1,000 words. 9 / 300 x 1000 = 30.

  5. The Principle of Accountability requires you to
    • count only the variant you are interested in
    • use only one speaker
    • ignore non-standard forms
    • count a variant against every occurrence of the variable
    Show answer

    count a variant against every occurrence of the variable. Essentials of Linguistics: count all possible occurrences.

  6. Why normalise frequencies?
    • To make the numbers bigger
    • Because texts differ in length, so raw counts cannot be compared
    • Because Pearson requires decimals
    • To remove non-standard forms
    Show answer

    Because texts differ in length, so raw counts cannot be compared. A longer text will contain more of almost everything.

  7. You found a pattern in two short texts. How should you present it in Section B?
    Show answer

    Cautiously, as a tentative finding, because a small sample may not be representative.. A small sample is a hint, not proof; acknowledging this is evaluative.

Practise with instant feedback

Sources

Last updated 8 November 2026. Spotted a mistake? Tell us.