We have had our model challenged.
Months of preparation, 303 people working, and a questionnaire designed so that our own model could have easily failed it. Below is what came out, including what was incorrect.
Research is not our primary service. Nevertheless, we have invested months in it. Not because we had to, but because we believe you should be able to verify what you're working with. We are building an ecosystem for development, for professionals who train people and organizations that want to advance people. That includes daring to expose your own model to the possibility that it might be incorrect.
What you find below.
What we investigated
whether behavior is directly related to what someone is missing, or if it goes through A story someone tells themselves.
What we saw
The direct connection disappears completely as soon as the narrative comes into the calculation. It held up under five different checks.
What it means
Directly changing behavior doesn't work. It's attached to a story that must be visible first.
Most behavioral models have never been tested.
A lot of money is spent annually on instruments that put people into boxes. Colors, letters, types. They feel recognizable, they sell well, and almost no one asks if they're accurate. A model that can never be wrong can also never tell you anything.
We are not testing a new foundation. The fact that people have three basic needs has been substantiated for decades. The fact that people give meaning to their own behavior and act accordingly, also. Those links are not up for debate. What INR adds is the connection between them. And that's what we tested.
How we worked.
Quantitative. Each participant received 36 statements: three layers, three needs, four statements per combination. Everything on a scale of 1 to 7.
The same structure for everyone
Each participant answered the same 36 statements, in the same order. No modifications, no exceptions.
Controls hidden in the list
Four built-in controls filtered out noise before anything was even calculated. Those who clicked through were eliminated.
Only calculate afterwards
The analysis looked at whether the relationship between need and behavior holds once the narrative is included.
The three layers of the INR Model form the core of the questionnaire. Each layer was measured separately, and only then were they linked to one another.
The I-layer
What you miss. Under pressure, you experience too little space, competence, or connection.
The N-layer
What you tell yourself. The story that forms around it, based on what you see in yourself.
The R-layer
What others see. The behavior that arises from that, intended as protection.
Scientific justification
This study does not test a new foundation, but a link between two foundations that have long been substantiated individually. The I-layer builds on research into basic human needs, in which autonomy, competence, and connectedness are central. The N-layer and the R-layer are based on the insight that people observe their own behavior, infer who they are from it, and then behave according to that story.
What INR adds is the connection: the assumption that the degree to which a need is met signals which narrative is emerging, and that behavior arising from that narrative subsequently confirms it. That is what has been tested here.
The analysis uses a mediation model with bootstrapping. Reliability is reported with both Cronbach's alpha and McDonald's omega. The outcome is replicated under five different cleaning rules to check if it depends on analytical choices.
Theoretical anchoring. The substantiation is based on established scientific work:
- Deci, E. L., & Ryan, R. M. (2000). Self-determination theory: Toward a motivational science of human.
- Bem, D. J. (1972). Self-Perception Theory, or how people observe their own behavior and infer who they are from it.
The narrative is the link.
The chain allowed itself to be tested. As soon as the narrative is included in the calculation, the direct link between what someone misses and how someone behaves disappears.
What you're missing
Under pressure, you experience too little space, competence, or connection.
What you tell yourself
A story forms around that. “I have to solve this myself.”
What others see
And that leads to behavior. You pull everything towards you and don't let go.
Behavior is not tied to need. It is tied to the story.
The connection between need and behavior is strong, but it's not a direct one. It runs through the narrative. That's precisely what our model predicted, and it could have gone differently.
Is the instrument reliable?
A measurement is only useful if it is consistent. These figures show whether the statements within each layer measure the same thing. The higher, the more stable.
| Layer | Alpha | Omega | What this means |
|---|---|---|---|
| I · what you miss | 0,87 | 0,89 | Good. The measurement is stable. |
| N. what you tell yourself | 0,82 | 0,86 | Good. It was once at 0.46 and has been rebuilt. |
| R · what others see | 0,81 | 0,86 | Good. The measurement is stable. |
A value above 0.80 is considered good. The fact that the N-layer was once at 0.46 and is now at 0.82 is not a minor detail. We found that weakness ourselves, named it, and rebuilt the premises.
Does it remain standing if you change the rules?
Researchers can manipulate their outcomes by choosing which participants to exclude. Therefore, we reran the analysis five times, each time with different rules.
| Rule | Participants | Direct connection | Through the narrative |
|---|---|---|---|
| Nobody left behind | 346 | 0,02 | 0,55 |
| Our rule | 303 | 0,00 | 0,51 |
| String | 175 | 0,04 | 0,51 |
| Get rid of only click-throughs | 341 | 0,02 | 0,54 |
| Very strict | 165 | 0,04 | 0,52 |
In all five variants, the direct connection remains practically zero, and everything goes through the story. The outcome does not depend on how we cleaned the data.
We are building safety nets within the test itself.
A questionnaire that everyone fills out primarily measures how fast people can click. That's why controls are hidden within it to filter out noise before we calculate anything.
The attention span
Somewhere in the list is a statement that literally asks you to click on a specific answer. Anyone who doesn't read the question will be caught out.
42 participants excludedThe clickers
Anyone who gives the exact same answer to all 36 statements is not filling in answers but clicking through. Those rows will be removed.
5 participants excludedThe mirror questions
Two statements ask the same thing, but phrased in reverse. Anyone who affirms both contradicts themselves.
Weighed as a signalSocial desirability
Two statements measure whether someone is pretending to be better than they are. We did not exclude anyone on that basis, but we do know who did it.
Measured, not filteredOut of 346 submissions, 43 were rejected. That's almost one in eight. We'd rather mention that than pretend everyone filled it out carefully. And that's precisely why the previous test matters: even if you just leave that 43 as is, the outcome doesn't change.
This is what 303 people filled in.
Each row is a participant. Each column is a statement, answered on a scale of 1 to 7. Below is a fragment, to show where the numbers come from.
| # | I_space | N_space | R_space | I_competent | N_competent | R_competent |
|---|---|---|---|---|---|---|
| 001 | 3 | 5 | 7 | 5 | 7 | 6 |
| 002 | 2 | 2 | 2 | 1 | 1 | 4 |
| 003 | 6 | 7 | 6 | 3 | 5 | 6 |
| 004 | 2 | 3 | 2 | 6 | 2 | 5 |
| 005 | 3 | 5 | 6 | 3 | 5 | 7 |
| 006 | 4 | 4 | 6 | 5 | 4 | 6 |
Look at participant 002: little sense of loss, little of a story, little protective behavior. And participant 003: a strong sense of loss, a strong narrative, strong behavior. That pattern, across 303 people, is what the analysis makes visible.
What this study doesn't say
- No cause and effect. We demonstrate coherence, not mechanism. The direction is theoretically substantiated, but not established in this study itself.
- No measurement over time. All three layers were measured at the same time. You cannot therefore follow a process over time.
- The behavior is self-reported. We measure what people say about their own behavior, not what others observe in them.
- This is less applicable to connectedness. There remains a direct link between lack and behavior. This is a limitation of our findings, and we prefer to mention it ourselves rather than have someone else discover it.
The model submitted itself to testing, and held firm.
In 303 people, behavior is not directly linked to what someone is missing, but to the narrative in between. That pattern remained under every test we applied. No evidence of cause and effect, but rather of a model that could have been refuted and was not.