Skip to content
CiderHQ
Search

Tasting guide

Structured sensory assessment

How is cider assessed when the answer has to hold up?

In short

By removing the taster from the result as far as possible. Controlled and identical serving conditions, samples presented blind and in randomised order, a fixed vocabulary agreed in advance, reference standards for the terms that matter, and each sample presented more than once so that a taster’s own consistency can be measured.

A trained panel is not a group of people with good palates. It is a group who have been calibrated against physical reference standards so that when four of them write "astringent" they mean the same sensation at roughly the same intensity.

The honest limitation is statistical. A panel of six untrained people tasting once produces an impression, not a result, and no amount of care in the protocol changes that.

Who this is for. A producer running quality assessment, somebody organising a judging, or a reader who wants to understand what a published sensory study actually did. (assumes a working practice · 6 steps)

Worth reading first
Where scale changes the answer

Full descriptive analysis with a trained panel is a research method with a real cost in people-hours. The intermediate practice — blind, randomised, replicated, with a fixed vocabulary — is achievable by a small producer and delivers most of the protection against self-deception.

The path

Each step says what happens, why it matters downstream, and what you are actually deciding. The links under a step are where the depth is.

  1. 1. Control everything that is not the sample

    Identical glasses, identical volumes, identical temperature, neutral lighting or coloured glasses where appearance would bias the palate, and no odour in the room. Samples coded so that nobody handling them knows which is which.

    Order is a control too. The first sample of a session is assessed differently from the fourth, so presentation order is randomised across tasters rather than being the same for everybody.

    Why it matters: Every uncontrolled variable becomes part of the result and is indistinguishable from it afterwards.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Washington State University cider research programme

      Washington State University Northwestern Washington Research and Extension Center · university · passage verified 2026-08-24

      The single most useful open cider dataset CiderHQ has found. The programme’s cultivar performance database gives juice chemistry, orchard behaviour, bloom and harvest timing and cider-maker tasting notes for 73 cultivars grown at one maritime site over fifteen years, and — unusually — publishes its classification thresholds alongside its figures, so the classification can be checked rather than taken on trust. It also states outright that its results diverge from the English ones. Read in full on 2026-08-24 and transcribed into `data/trials/wsu-mount-vernon.ts`; four published figures were found to be impossible and are withheld there with their reasons.

  2. 2. Fix the vocabulary, and anchor it to something physical

    A panel that has not agreed its terms is measuring the terms rather than the cider. The vocabulary is fixed in advance, each term is defined, and the ones that matter are anchored to a reference standard — a physical sample that demonstrates the sensation, which every taster smells or tastes before scoring.

    This is where published cider lexicons earn their keep. A lexicon developed with a trained panel is not a list of pleasant words; it is a set of terms that have been shown to be discriminable and reproducible across tasters.

    Why it matters: Without anchoring, two tasters agreeing on a word is not evidence that they perceived the same thing.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Washington State University cider research programme

      Washington State University Northwestern Washington Research and Extension Center · university · passage verified 2026-08-24

      The single most useful open cider dataset CiderHQ has found. The programme’s cultivar performance database gives juice chemistry, orchard behaviour, bloom and harvest timing and cider-maker tasting notes for 73 cultivars grown at one maritime site over fifteen years, and — unusually — publishes its classification thresholds alongside its figures, so the classification can be checked rather than taken on trust. It also states outright that its results diverge from the English ones. Read in full on 2026-08-24 and transcribed into `data/trials/wsu-mount-vernon.ts`; four published figures were found to be impossible and are withheld there with their reasons.

    • BJCP Beer Style Guidelines — cider and perry categories

      Beer Judge Certification Program · classification framework · retrieved 2026-08-24

      Opened on 2026-08-24, and the edition matters. CiderHQ registered the 2015 cider guidelines; BJCP has since published a 2025 cider edition, and the old URL now redirects into the current guidelines index. A competition style guide is a dated framework rather than a standing definition, and citing the wrong edition of one is the same class of error as citing a superseded specification. CiderHQ treats BJCP as one competition framework among several and never as the universal cider taxonomy.

  3. 3. Present each sample more than once

    A sample presented twice within a session, without the taster knowing, measures that taster’s own reproducibility. A taster who scores the same cider markedly differently on the two presentations is not contributing signal, and the only way to discover that is to test it.

    Replication also separates panel disagreement from panel noise. Four tasters who disagree consistently are detecting something real about the sample; four who disagree differently each time are not detecting anything.

    Why it matters: It is the only internal check available. Without it, a panel result cannot be distinguished from a set of guesses that happened to cluster.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Washington State University cider research programme

      Washington State University Northwestern Washington Research and Extension Center · university · passage verified 2026-08-24

      The single most useful open cider dataset CiderHQ has found. The programme’s cultivar performance database gives juice chemistry, orchard behaviour, bloom and harvest timing and cider-maker tasting notes for 73 cultivars grown at one maritime site over fifteen years, and — unusually — publishes its classification thresholds alongside its figures, so the classification can be checked rather than taken on trust. It also states outright that its results diverge from the English ones. Read in full on 2026-08-24 and transcribed into `data/trials/wsu-mount-vernon.ts`; four published figures were found to be impossible and are withheld there with their reasons.

    • Aroma Release and Consumer Perception During Cider Consumption

      Chen, Zhao, Jiang, Liu, Cao, Ye, Mao and He, Foods 14(6):1005 · peer-reviewed literature · passage verified 2026-08-24 · covers 2025

      Open access; read in full on 2026-08-24. Registered for one finding that changes how every descriptor page on this site should be read: the aroma compounds that correlate with a perceived attribute are not the same set when the cider is sniffed as when it is in the mouth. Five commercial ciders, a trained panel and a consumer panel, with the headspace sampled both ways. A compound present above its threshold in the glass was in one case detectable retronasally in only one of the five samples.

  4. 4. Decide which question you are asking before choosing the method

    There are two broadly different things a panel can be asked to do, and they need different designs. A difference test asks whether two samples are distinguishable at all — the triangle test, where a taster receives three samples of which two are identical and has to identify the odd one, is the classic form. It answers a yes-or-no question with a known probability of being right by chance, which is what makes its statistics tractable.

    A descriptive analysis asks what the differences consist of, and requires a trained panel, an agreed vocabulary, intensity scales and considerably more time. It produces a profile rather than a verdict.

    The mistake worth avoiding is running a descriptive analysis to answer a difference question. If the practical decision is "does the new yeast change the cider", a triangle test with enough replicates answers it cheaply and defensibly, and a panel writing free-form notes does not answer it at all.

    Why it matters: Most production sensory questions are difference questions wearing descriptive clothes, and difference tests are far cheaper to run and far harder to argue with.

    What you are deciding

    • Difference or description. "Has anything changed" is the first; "what changed" is the second, and it is only worth asking once the first has been answered.
    • How many tasters and how many replicates, which for a difference test is a statistical question with a definite answer rather than a matter of what is convenient.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Washington State University cider research programme

      Washington State University Northwestern Washington Research and Extension Center · university · passage verified 2026-08-24

      The single most useful open cider dataset CiderHQ has found. The programme’s cultivar performance database gives juice chemistry, orchard behaviour, bloom and harvest timing and cider-maker tasting notes for 73 cultivars grown at one maritime site over fifteen years, and — unusually — publishes its classification thresholds alongside its figures, so the classification can be checked rather than taken on trust. It also states outright that its results diverge from the English ones. Read in full on 2026-08-24 and transcribed into `data/trials/wsu-mount-vernon.ts`; four published figures were found to be impossible and are withheld there with their reasons.

    • Aroma Release and Consumer Perception During Cider Consumption

      Chen, Zhao, Jiang, Liu, Cao, Ye, Mao and He, Foods 14(6):1005 · peer-reviewed literature · passage verified 2026-08-24 · covers 2025

      Open access; read in full on 2026-08-24. Registered for one finding that changes how every descriptor page on this site should be read: the aroma compounds that correlate with a perceived attribute are not the same set when the cider is sniffed as when it is in the mouth. Five commercial ciders, a trained panel and a consumer panel, with the headspace sampled both ways. A compound present above its threshold in the glass was in one case detectable retronasally in only one of the five samples.

  5. 5. Remember that thresholds vary between people by orders of magnitude

    Detection thresholds for individual compounds differ enormously between individuals, and for some compounds a proportion of the population cannot detect them at all at any relevant concentration. A panellist who reports nothing may be reporting accurately about their own perception.

    This has a direct practical consequence for fault assessment. A cider carrying a compound at a level some tasters detect and others do not will produce a genuinely split panel, and the split is a finding rather than a failure of the panel.

    Why it matters: It is why a single taster’s failure to detect a fault is weak evidence that it is absent, and why fault decisions in production are not made by one person.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Aroma Release and Consumer Perception During Cider Consumption

      Chen, Zhao, Jiang, Liu, Cao, Ye, Mao and He, Foods 14(6):1005 · peer-reviewed literature · passage verified 2026-08-24 · covers 2025

      Open access; read in full on 2026-08-24. Registered for one finding that changes how every descriptor page on this site should be read: the aroma compounds that correlate with a perceived attribute are not the same set when the cider is sniffed as when it is in the mouth. Five commercial ciders, a trained panel and a consumer panel, with the headspace sampled both ways. A compound present above its threshold in the glass was in one case detectable retronasally in only one of the five samples.

    • Peer-reviewed literature on apple phenolics and cider sensory perception

      Various journals · peer-reviewed literature · registered as competent for this subject

      Registered as a class rather than as one paper, because the mechanisms CiderHQ describes — tannin chain length driving the split between bitterness and astringency, salivary protein precipitation, enzymatic browning — are established across many studies rather than resting on any single one. Individual papers are cited where a specific number is quoted.

  6. 6. State what the assessment cannot conclude

    A small panel can reliably answer whether two samples differ. It answers "by how much" far less reliably and "which is better" not at all, since preference is not a sensory measurement.

    A published sensory result should carry its panel size, its training status, its replication and its statistics. A result reported without them is an impression with a method section attached, and this site treats it accordingly.

    The limitation that matters most for cider specifically is that the panel-trained lexicons which exist were developed on particular sets of ciders in particular places. A lexicon built on Pacific Northwest fruit is a real instrument and it is not automatically the right instrument for Asturian sidra natural, whose characteristic volatile profile would be scored as faulty by a vocabulary that never had to accommodate it. Applying a lexicon outside the material it was built on is a methodological choice and should be declared as one.

    Why it matters: Sensory data is the easiest kind of evidence to over-read, and the over-reading usually happens at the point where a difference becomes a ranking.

    Go deeper

    Sources

    What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.

    • Washington State University cider research programme

      Washington State University Northwestern Washington Research and Extension Center · university · passage verified 2026-08-24

      The single most useful open cider dataset CiderHQ has found. The programme’s cultivar performance database gives juice chemistry, orchard behaviour, bloom and harvest timing and cider-maker tasting notes for 73 cultivars grown at one maritime site over fifteen years, and — unusually — publishes its classification thresholds alongside its figures, so the classification can be checked rather than taken on trust. It also states outright that its results diverge from the English ones. Read in full on 2026-08-24 and transcribed into `data/trials/wsu-mount-vernon.ts`; four published figures were found to be impossible and are withheld there with their reasons.

    • Aroma Release and Consumer Perception During Cider Consumption

      Chen, Zhao, Jiang, Liu, Cao, Ye, Mao and He, Foods 14(6):1005 · peer-reviewed literature · passage verified 2026-08-24 · covers 2025

      Open access; read in full on 2026-08-24. Registered for one finding that changes how every descriptor page on this site should be read: the aroma compounds that correlate with a perceived attribute are not the same set when the cider is sniffed as when it is in the mouth. Five commercial ciders, a trained panel and a consumer panel, with the headspace sampled both ways. A compound present above its threshold in the glass was in one case detectable retronasally in only one of the five samples.

Where to go next

Guides that follow on from this one.

Questions this answers

Shown so you can tell whether this is the guide you wanted.

Where to go next

Sources

What this page rests on. Where a source is marked as registered rather than read, CiderHQ is recording that the body is authoritative on the subject without claiming to have worked through the document itself. See our evidence policy for what each state means.