Interpreting Histograms: Shape, Skewness, Gaps, and Outliers

Welcome back. In the last lesson, you turned grouped data into a frequency table and learned that relative and cumulative frequency answer different questions. A histogram is the visual version of that grouped table: each bar represents a class interval, and its height shows how common values in that interval are.

Now the goal is to look past individual bar heights and describe the overall pattern. By the end of this lesson, you will be able to write a clear histogram description that identifies its shape, skewness, gaps, likely outliers, and whether it is approximately normal.

---

Read the overall pattern, not one bar at a time

A histogram displays quantitative data divided into numerical intervals, called classes or bins. Unlike a bar chart for categories, its bars normally touch because the number line is continuous: an interval such as 50–59 sits beside 60–69.

To interpret one, scan from left to right and ask:

1. Where are most observations concentrated?
2. How many main peaks or humps are there?
3. Do the two sides look balanced, or does one side stretch farther?
4. Are there empty intervals, separate clusters, or isolated bars?
5. Does the whole pattern resemble one smooth, balanced mound?

A useful description reports the broad pattern rather than treating every small rise or drop as meaningful. Histograms, especially those based on small samples, can look a little uneven simply because of random variation.

Read the short sections below for the vocabulary that supports this process.

---

Shape: peaks, symmetry, and skewness

Start with the number of peaks

A mode is a peak or major concentration of observations. Describe a histogram as:

• Unimodal if it has one clear main peak.
• Bimodal if it has two clearly separated peaks.
• Multimodal if it has more than two main peaks.
• Approximately uniform if the bars are fairly similar in height across the range.

Do not call every slightly taller bar a separate mode. For a true bimodal distribution, there should be two noticeable humps with a dip between them.

Two peaks can be a clue that the data combine different groups. For example, combining the heights of a group of children and a group of adults could produce two clusters. A histogram cannot prove why this happened, but it can tell you that investigating groups may be worthwhile.

Then judge balance

A distribution is roughly symmetric when its left and right sides are approximate mirror images around the center. The tails taper off by similar amounts on both sides.

A distribution is skewed when one tail extends farther than the other:

• Skewed right: the longer tail points toward larger values, on the right.
• Skewed left: the longer tail points toward smaller values, on the left.

The most reliable rule is:

Name skewness for the direction of the long tail, not the direction of the peak.

In a right-skewed histogram, the bulk of the data is often toward the left or middle, while a smaller number of unusually high values extend the graph to the right. Income data often have this pattern: many moderate incomes and a few very large ones.

In a left-skewed histogram, the bulk is toward the right or middle, while a smaller number of unusually low values create a tail to the left. Scores on an easy test can sometimes be left-skewed: many high scores and a smaller number of low scores.

This video gives quick visual examples of the labels.

A numerical clue can sometimes support what you see later in the course:

Visual pattern	Typical relationship
Roughly symmetric	Mean and median are approximately equal
Skewed right	Mean is usually greater than the median
Skewed left	Mean is usually less than the median

This is a clue, not a substitute for inspecting the histogram. Extreme values pull the mean toward the long tail.

---

Gaps and likely outliers

A gap is an interval of values with no observations. In a histogram, it appears as one or more empty spaces between bars.

A gap can mean different things:

• There may truly be no observed values in that range.
• There may be two separate clusters in the data.
• It can help reveal that an isolated value is unusual.

An outlier is an observation unusually far from the overall pattern. On a histogram, you should normally call it a possible or likely outlier, because the graph groups data into intervals. A short isolated bar might represent one value, but it could represent several values within that interval.

Use this distinction carefully:

• A long thin tail has several low bars gradually stretching outward. That suggests skewness.
• An isolated bar after a visible gap suggests a likely outlier.
• A single extreme bar can make a distribution look skewed, even when the main cluster is reasonably symmetric.

For instance, if nearly all test scores lie between 60 and 90 but one score lies near 20 with an empty interval between 20 and 60, the low score is a likely outlier. A careful description would be:

The main distribution is unimodal and roughly symmetric, with a possible low outlier near 20.

That wording is better than simply calling the entire distribution “left-skewed,” because it distinguishes the main pattern from one unusual observation.

If raw data are available, investigate an apparent outlier before deleting or ignoring it. It could be a data-entry error, a measurement error, or a genuine but important observation.

---

When is a histogram approximately normal?

A normal distribution has a particular idealized shape. It is:

• unimodal;
• symmetric;
• mound- or bell-shaped;
• highest near the center;
• gradually tapering in both directions;
• without strong gaps, multiple peaks, or extreme isolated values.

Real data do not need to look perfectly smooth to be approximately normal. A sample histogram may be a little jagged, particularly with fewer observations. The question is whether its general outline fits a single balanced mound.

What rules out approximate normality?

These features are evidence against a normal model:

Histogram feature	Why it is not approximately normal
Strong right or left skew	A normal distribution is symmetric.
Two clear peaks	A normal distribution has one peak.
Several clusters	The data do not form one continuous mound.
A substantial gap	A normal curve tapers smoothly rather than breaking apart.
A clearly isolated extreme value	The outlier may distort the overall pattern.
Nearly flat bars throughout	This is closer to uniform than normal.

Be careful with a common overstatement: symmetric does not automatically mean normal. A distribution can be symmetric and bimodal, flat, or otherwise non-normal. “Normal” requires the combination of one central peak, rough symmetry, and a smooth bell-like pattern.

At this stage, describe normality visually. Later, you will use normal quantile plots and numerical ideas to assess it more formally.

---

A complete description checklist

On a test or assignment, a dependable order is:

1. Shape and modality: unimodal, bimodal, multimodal, or uniform.
2. Symmetry or skewness: roughly symmetric, skewed right, or skewed left.
3. Gaps or clusters: state where they occur, if present.
4. Likely outliers: state their approximate location and why they stand out.
5. Approximate normality: give a conclusion and connect it to the visible evidence.

Use approximate values from the horizontal axis when you can. For example:

The distribution is unimodal and moderately skewed right, with most observations between 20 and 50 and a long tail extending toward 100. There are no obvious gaps, but the highest values may be possible outliers. Because the distribution is skewed rather than symmetric, it is not approximately normal.

That is stronger than writing only “right-skewed,” because it supplies the evidence and addresses the full pattern.

---

Reading the four histograms

Apply the checklist to the panels in the image.

Histogram A

A has one dominant peak, so it is unimodal. The bars taper on both sides of that peak and there is no visible gap or separated bar. Its shape looks roughly symmetric, although not perfectly so.

A reasonable description is:

Histogram A is unimodal and roughly symmetric, with no apparent gaps or likely outliers. It has a plausible approximately normal shape.

“Approximately” matters. A histogram is not expected to trace a perfect bell curve.

Histogram B

B is also unimodal. It has a central cluster, but its bars stretch somewhat farther toward the larger values on the right. The graph is therefore at least slightly skewed right, rather than perfectly symmetric.

A sound conclusion is:

Histogram B is unimodal with a mild right tail and no obvious gap or isolated outlier. It is less convincingly normal than A because the symmetry is weaker.

With graphs like B, avoid pretending that the judgment is exact. “Slightly right-skewed” or “roughly symmetric with a slight right tail” is often more defensible than an overconfident label.

Histogram C

C covers a much broader range and has an uneven, extended pattern toward higher values. It does not make one smooth, balanced mound. The higher-value region is relatively sparse, producing a rightward tail.

A useful description is:

Histogram C is broad and irregular, with a general right-skewed pattern. It is not approximately normal because it lacks a single, symmetric bell-shaped form.

Histogram D

D has a strong central cluster, but there is also a small separated bar far to the right. The empty region before that last bar is crucial: it suggests the high value or values are likely outliers.

A careful description is:

Histogram D has a unimodal, roughly symmetric main cluster, plus a likely high outlier near the far right after a noticeable gap. The full distribution should not be treated as clearly normal because of that separated extreme value.

Notice the difference between saying “the data are right-skewed” and describing the main body plus an outlier. The latter is more informative and often more accurate.

---

Key takeaways

A histogram description should tell the story of the distribution:

• Identify its number of meaningful peaks.
• Describe it as roughly symmetric or name the direction of its long tail.
• Look for gaps, clusters, and separated bars.
• Call visually isolated values possible or likely outliers.
• Call a distribution approximately normal only when it is broadly unimodal, symmetric, and bell-shaped, without striking gaps or outliers.

The grouped frequency tables from the last two lessons help you construct a histogram; the histogram then makes patterns that are hard to see in a table much easier to recognize.

This completes the visual-summary work for Week 1. Next, you will begin the Excel essentials module by learning to enter and organize a small dataset using labels, values, cell references, and basic formulas.