TOEFL Speaking Scores by First Language: What 200 Practice Records Show

The highest group average in this sample is 3.55/5. The lowest is 2.43/5. Yet the score range within every language group is wider than the gap between those averages.

That is the most useful finding in this comparison of Chicago Interview practice scores on My Speaking Score. Group averages differ, but they conceal substantial variation within each first-language group.

The interactive chart below shows all 200 score records, grouped by first language. Each group contains exactly 20 records. You can compare individual scores, averages, and medians, then see what changes when zero scores are excluded.

Explore the Chicago Interview scores

Move over a row or tap it to inspect a score. The group selector provides keyboard access to group summaries. Overlapping equal scores are stacked. Changing the zero-score filter recalculates the chart and table without changing the findings reported below.

What the sample shows

Hindi has the highest average in this file, at 3.545/5 before rounding. Urdu has the lowest, at 2.430/5. The difference is 1.115 points.

But the narrowest within-group score range is 1.8 points: Hindi scores run from 2.5 to 4.3. Every other group has an even wider observed range. This means the spread within each group exceeds the gap between the highest and lowest group averages.

All ten groups include scores between 2.9 and 3.5. The distributions overlap, so a first-language label alone cannot tell you where a particular record falls.

Across all 200 records, the average is 3.007/5 and the median is 3.20/5. These are descriptive results for this file. They are not population estimates for TOEFL test takers.

The medians tell a less spread-out story

The average uses every score. The median is the midpoint of the ordered scores, averaging the two central values when the count is even.

In this sample, group medians range from 2.95 to 3.50, a gap of 0.55 points. That is smaller than the 1.115-point gap between group averages. Lower scores pull some averages further down while their medians remain near 3.

For example, the Urdu group has an average of 2.43 and a median of 2.95. The Mandarin group has an average of 3.15 and a median of 3.35. Neither statistic describes the entire distribution, which is why the chart keeps the individual records visible.

The group summaries

Averages use three decimal places to preserve the calculation. These are descriptive summaries, not rankings of language communities. The table follows the chart controls.

Three zero scores affect the comparison

The file contains three zero scores: one in the Arabic group and two in the Urdu group. All zeros are included in the main analysis.

Excluding them changes the Arabic average from 2.755 to 2.900 and the Urdu average from 2.430 to 2.700. The chart lets you inspect that sensitivity directly.

The export does not identify why any record received zero. We cannot classify these as off-topic responses, missing recordings, or technical failures from this file. Excluding them is an alternative view, not a correction to the data.

What this comparison cannot explain

The source contains only a first-language label and an Interview score for each row. It does not include proficiency level, preparation history, individual dimension scores, response transcripts, dates, or learner identifiers.

We therefore cannot determine whether records represent distinct learners or repeat attempts. We also cannot establish how the 20 records per group were selected. Equal group sizes make the comparison balanced by count, but they do not establish representative sampling.

The differences do not establish that first language caused a score difference. They also do not identify fluency, pronunciation, grammar, organization, or task coverage as the explanation for any group’s results. Those questions require additional evidence.

What learners should take from the chart

Your first-language group average is not a diagnosis of your speaking. This sample shows substantial variation among records carrying the same language label.

For a useful practice decision, examine your own response: whether you answered the question, how clearly the listener can understand you, where delivery breaks down, and whether you developed your ideas. Those observations can guide a specific next step.

Methodology

Source: chicago_interview_scores_by_l1.csv, supplied for this analysis. We grouped all 200 rows by native_language and summarized interview_score_unrounded. There are 10 groups with 20 rows each. Scores in the file range from 0.0 to 4.5 on the zero-to-five practice scale. No rows were removed from the main analysis.

The chart uses a full zero-to-five axis. Each dot represents one source row. Equal scores are stacked for visibility. Average and median markers are calculated from the currently included records. No significance testing, causal modeling, or population weighting was performed.

Frequently asked questions

Find your own next step

Take the free Chicago test at www.myspeakingscore.com. Use your own response and feedback to identify what to practice next.

My Speaking Score practice scores are not official TOEFL iBT® scores.