G2 Science K223, K224 and K225 often asks how an outcome changes as a factor changes. One treatment level can show a difference; several levels can reveal the shape of the response: linear, thresholded, saturating, optimal, declining or otherwise conditional.
This one-hundred-and-thirty-fifth Learner’s Guide develops dose–response shape reasoning in a broad school-science sense: “dose” means the level of an input or factor, not necessarily a medicine. It extends Vol 0107 Interaction Effects and Vol 0131 Bottleneck Reasoning.
The response-shape protocol
Measure or compare several input levels across a justified range. Plot or tabulate input against response. Ask whether the effect is proportional, thresholded, saturating, peaked or changing direction. Then connect the shape to a plausible mechanism and the limits of the design.
1. light intensity response
For light intensity response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
2. carbon dioxide response
For carbon dioxide response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
3. temperature response over limited range
For temperature response over limited range, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
4. enzyme concentration response
For enzyme concentration response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
5. substrate concentration response
For substrate concentration response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
6. reactant concentration response
For reactant concentration response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
7. surface area response
For surface area response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
8. catalyst amount response in given context
For catalyst amount response in given context, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
9. water availability response
For water availability response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
10. nutrient availability response
For nutrient availability response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
11. oxygen availability response
For oxygen availability response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
12. exercise intensity response
For exercise intensity response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
13. stimulus intensity response
For stimulus intensity response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
14. load-extension response in valid range
For load-extension response in valid range, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
15. current-heating response
For current-heating response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
16. voltage-current response under stated component behaviour
For voltage-current response under stated component behaviour, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
17. magnetic distance response
For magnetic distance response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
18. pollutant concentration response
For pollutant concentration response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
19. exposure duration response
For exposure duration response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
20. dose-like concentration×time idea in context
For dose-like concentration×time idea in context, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
21. resource availability
For resource availability, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
22. population density effect
For population density effect, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
23. predator density effect
For predator density effect, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
24. prey density effect
For prey density effect, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
25. treatment level
For treatment level, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
26. pH response
For pH response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
27. salinity-like concentration context
For salinity-like concentration context, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
28. humidity response
For humidity response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
29. airflow response
For airflow response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
30. distance from source
For distance from source, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
31. signal strength
For signal strength, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
32. sensor input
For sensor input, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
33. linear response
For linear response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
34. proportional response
For proportional response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
35. sublinear response
For sublinear response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
36. superlinear response
For superlinear response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
37. threshold response
For threshold response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
38. saturation response
For saturation response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
39. plateau
For plateau, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
40. U-shaped response
For U-shaped response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
41. inverted-U response
For inverted-U response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
42. optimum
For optimum, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
43. decline after optimum
For decline after optimum, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
44. toxic/excess range where context states
For toxic/excess range where context states, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
45. no-effect range
For no-effect range, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
46. minimum effective level
For minimum effective level, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
47. maximum useful level
For maximum useful level, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
48. ceiling effect
For ceiling effect, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
49. floor effect
For floor effect, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
50. baseline response
For baseline response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
51. control zero/reference
For control zero/reference, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
52. low dose
For low dose, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
53. medium dose
For medium dose, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
54. high dose
For high dose, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
55. more levels
For more levels, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
56. single-level limitation
For single-level limitation, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
57. range too narrow
For range too narrow, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
58. range too wide
For range too wide, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
59. spacing of levels
For spacing of levels, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
60. replicates at each level
For replicates at each level, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
61. mean response
For mean response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
62. variation around mean
For variation around mean, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
63. outlier
For outlier, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
64. response lag
For response lag, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
65. baseline drift
For baseline drift, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
66. interaction with second factor
For interaction with second factor, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
67. path dependence
For path dependence, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
68. bottleneck shift
For bottleneck shift, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
69. limiting factor
For limiting factor, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
70. mechanism predicts shape
For mechanism predicts shape, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
71. graph shape
For graph shape, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
72. table shape
For table shape, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
73. slope changes
For slope changes, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
74. turning point
For turning point, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
75. breakpoint
For breakpoint, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
76. interpolation
For interpolation, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
77. extrapolation
For extrapolation, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
78. model range
For model range, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
79. causal inference
For causal inference, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
80. correlation only
For correlation only, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
81. reverse causation less plausible with controlled manipulation
For reverse causation less plausible with controlled manipulation, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
82. confounder
For confounder, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
83. measurement ceiling
For measurement ceiling, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
84. detector floor
For detector floor, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
85. normalise response
For normalise response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
86. rate response
For rate response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
87. amount response
For amount response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
88. time-course response
For time-course response, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
89. same endpoint different curve
For same endpoint different curve, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
90. same curve different scale
For same curve different scale, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
91. dose-response final rule
For dose-response final rule, predict the response at low, medium and high input before looking at data. Then compare the observed shape with the prediction and identify where the model succeeds or fails.
For practice, add one extra input level chosen specifically to distinguish two competing shapes. A good new level is not merely “more data”; it sits where the models make different predictions, such as near a threshold, plateau or turning point.
Do not generalise beyond the tested range. A rising trend at low levels does not prove indefinite increase. Saturation, limiting factors, interactions or harmful/excess conditions may change the relationship outside the observed interval.
Response shape can diagnose mechanism
A plateau can suggest saturation or a new bottleneck; a threshold suggests a minimum condition; a peak followed by decline suggests an optimum or competing effect. Shape alone does not prove the mechanism, but it creates discriminating predictions.
Links
Use the Science Hub, Vol 0123 Path Dependence, Vol 0103 Model-Validity Boundaries, the Examination Craft hub and the PSLE Learner’s Guide.
Official-source discipline
For the current 2027 SEC G2 school-candidate framework, use the official SEAB G2 syllabus directory and linked K223–K225 Science syllabuses. Dose–response shape is used here as an eduKateSengkang experimental-reasoning framework, not as an additional SEAB syllabus label.
Final rule
Do not ask only whether more input gives more output. Ask what shape connects input to response, where that shape changes, and what new measurement would distinguish the competing explanations.
