Consistency Analysis

The Consistency Analysis measures how repeatably a participant produces the same IPA target within a session. It reports token accuracy, the number of distinct IPA Actual forms, and whether repeated productions are accurate or inaccurate and consistent or inconsistent.

Use this analysis to distinguish a stable production pattern from variable productions of the same target.

Before you begin

Each result requires aligned IPA Target and IPA Actual words. Check word boundaries, syllabification, and phone alignment, and assign a speaker to each record.

Within each participant and session, the analysis groups results that have the same IPA Target. A group is a word unit only when it contains at least two tokens. A target that occurs once is excluded.

Note:

The grouping key is the IPA Target form, not Orthography. If different orthographic words share the same IPA Target in one session, their tokens belong to the same word unit; the report displays the Orthography from the first token in that group.

Parameters

The analysis opens at the Parameters step. Its principal controls are:

Table 1. Consistency Analysis parameters
Parameter	Default	Effect
Report Title	Consistency Analysis	Sets the title of the generated report.
Combine results for all selected participants	Off	Combines selected participants instead of producing a separate participant section for each one. Results remain divided by session.
Diacritic Options	Compare diacritics	Controls whether all or selected diacritics are ignored for accuracy, distinct-type, and consistency comparisons.
Ignore Distortion	Off	When selected, an IPA Actual phone marked with the distorted diacritic is treated as correct for its corresponding target position. This changes accuracy but does not merge distinct production types.
Search by	Word	Searches aligned IPA Target and IPA Actual word domains. Phon automatically includes the aligned Orthography in the result.
Aligned Phones	Empty IPA Target and IPA Actual Phonex matchers	Optionally restricts the aligned words included in the analysis.

The remaining expandable controls are the standard tier, word, syllable, additional-tier-data, and participant filters. See Common Query Parameters, Diacritic Options, and Aligned Phones.

Figure 1. Consistency Analysis Parameters

Accuracy and consistency categories

A token is accurate when its IPA Actual has the same element sequence as its IPA Target after applying the selected diacritic options. The transcripts must have the same number of elements. Ignore Distortion modifies this comparison as described in the parameters.

A word unit is accurate when at least one token is accurate. It is consistent when every IPA Actual form in the unit is the same after applying the selected diacritic options. These independent tests place the unit in one category:

Table 2. Consistency categories
Code	Category	Rule
`A & C`	Accurate and Consistent	At least one token is accurate and the unit has one production type.
`I & C`	Inaccurate and Consistent	No token is accurate and the unit has one production type.
`A & I`	Accurate and Inconsistent	At least one token is accurate and the unit has more than one production type.
`I & I`	Inaccurate and Inconsistent	No token is accurate and the unit has more than one production type.

Word-unit measures

Count: The number of tokens in the word unit.
Accuracy (%): The number of accurate tokens divided by Count, multiplied by 100.
# Types: The number of distinct IPA Actual forms after applying the selected diacritic options.
TTR (%): The type-token ratio: # Types / Count × 100. Five identical tokens have one type and a TTR of 20%; five different types across five tokens have a TTR of 100%.

Report output

The report begins with a Readme and then organizes results by participant and session:

Table 3. Consistency Analysis report tables
Section or table	Contents
Participant Summary	One row per session, with the number of word units, mean word-unit accuracy, percentage in each category, mean number of types, and mean TTR.
Session Summary	One row per word unit, ordered by Orthography and IPA Target. It lists Count, Accuracy, the four category flags, # Types, and TTR. A final row reports unweighted means across the session's word units; the category columns become the percentage of units in each category.
Four category sections	Accurate and Consistent, Inaccurate and Consistent, Accurate and Inconsistent, and Inaccurate and Inconsistent. Each contains a token listing for every word unit in that category, ordered by Orthography. An empty category displays No results.

Each word-unit listing is titled with its Orthography and IPA Target and includes the source record, IPA Target, IPA Actual, Orthography, word index, and token-level accuracy. Select a result in the report to open its source record.