The "Magic Number" in Human Memory Capacity

Human Capacity Limits Process Industry

Recently, I wrote an article about a study from Shiraz University of Medical Sciences and their research on the topic of industrial operator sleepiness. Yokogawa has been conducting its own research in this area, specifically in the area of operator mental capacity. The purpose that drives this research is the need to understand the typical upper limit of information that can be fed to our operators at any time to ensure safe and efficient operations 24/7.

The industry as a whole is well aware that people have a limit in terms of what they can handle, but do we know what those limits are? Do these limits change throughout a shift? What is the associated impact on Commission and Omission errors?

Understanding some of these limits has been the core focus of our study so that we can support our customers to maximize the effectiveness of their operations, without pushing their operations teams into dangerous zones in terms of their how they are operating their facility.

Analysis Method

At this stage in our study we have been conducting analysis in a fixed environment as part of a standardized training course, meaning that all participants will have arrived at the location at the same time, and they would have conducted the same work and mental loading in the lead up to the testing.

At this stage we have been working to validate the efforts of George A. Miller from Princeton University and his research in a similar area that identified the “Magic Number”.

In our closed environment we conducted a structured buildup in which we gradually requested the attendees to memorize slightly longer chunks of information, including repetition of the same number of chunks but with different items. After reaching a certain point in the testing we then changed the format slightly by adding background noise, switching from letters to number, and too objects.

The test was conducted by showing the items for a limited time period, upon the items disappearing from the screen the participants were then able to start noting the items that they could remember. A test run was conducted at the beginning so that the attendees could have an expectation of what would occur.

As per the table 1 below:

Table 1: Test Structure

Table 1: Test Structure


At this stage we have a sample of 59 people; this will be increased over the course of future sessions in various regions of the world.

If you are interested in more details about the research please attend one of our upcoming workshops, of which details can be gained by contacting

First, let me get the high level information out of the way. The existence of a “Magic Number” identified by George Miller is definitely true, and this has proven itself in our results as seen in Chart 1.

Chart 1: Human Memory Capacity - Results of memory testing.

As you can see all participants hover just under the 7 target; nearly always in range. But very rarely do they exceed the target, except when we switch to memorizing objects.

When we start to drill down into these results we start to see more interesting results. As you can see below as we reach the target range (between 5 and 9) we see a slight decrease in the accuracy of the memory test results; as we continue to push higher into the range we see a continued deterioration of the memory test results.

So, whilst the memory capacity is approximately 7 items, the more memory load or mental pressure that we apply the actual memory capacity reduces and we end up with a situation wherein the operator is actually reaching a limit of between 4 to 6, because we are overwhelming them with items.


Chart 2: Human Memory Capacity - Percent in Range


In test #2-8 you will notice that we added background noise. We were particularly interested to understand the impact of distractions that would occur in the background and if these would have a noticeable negative impact of the memory capacity.

We conducted this by repeating a test that had been conducted earlier; but with 2 test leaders loudly yelling and reading random letters and numbers into the room during the recall and noting phase of the test

As you can see in the chart below, there was a noticeable deterioration in the memory capacity as a result of the background noise.


Chart 3: Human Memory Capacity - Impact of Distraction


Finally, we wanted to understand the impact of moving from a single-encoding method to a dual-encoding method, via the utilization of chunks. In this case we showed objects to the participants rather than letters or numbers. In addition you will notice that we overloaded them with 20 items to remember, in our next test run we will be adding a 10 item test.

Chart 4: Human Memory Capacity - Are object easier to remember?

What is significant here is that the move to dual-encoding of our information chunks has a significantly positive impact on information recall rates. We can see that the average has jumped to the upper end of our target range, with a sizable number of participants at or above the 9 item level.


From this data we can extract a number of key learning points:

  1. We have proven that we need to keep the number of information points being used by an operator at any time to within 7, and at the maximum 9. By information points we do not mean those purely shown to him from the operator console. This is all information from all sources; Console, Radio, email, Operator Logbook, Production Schedule. All information that he/she needs to make the right decision should be contained within 5 to 9 items of information.
  2. The harder we push the operator to search, hunt and wade through data points and information to make a decision, the less information he will actually remember well…leading to stress, inefficiency and potential for errors. This impact will be magnified as we saw in our sleepiness research.
  3. The more effort we can put into understanding the operators via deep Work Domain Analysis so that we can pre-visualize information in a manner that maps to the operators mental model. Resulting in the operator using a lower volume of high quality information that takes him straight or closer to the answer. Exactly as we are doing as part of our Advanced Decision Support and Advanced Operating Graphics services.


As I have written about in several articles in the past, it is very critical that we try to understand our operations by understanding our operators, from their point of view. Then, when implementing technologies, we focus on implementation in a manner that matches with the manner in which the operators think, and towards the goals that they need to achieve. Too often there is a focus on adding new and extra information/data.

As we have seen in this research so far there is a strong need to ensure that we do not overwhelm our operators with too much data; we should be focusing on small volumes of high value information. This needs to be conducted from a holistic perspective that covers all sources of information/data that the operators gather or receive.

Whilst I do not have hard data to prove it yet, I strongly believe that companies that can realize this will see a jump in production performance, quality and associated revenues; that is before considering the HSSE and staff retention topics.

For those who are interested to learn more about our research in this area please contact us directly at; also our European members will be holding an Advanced Decision Support workshop in the UK on October 11th 2017 of which you can register here.