MCAT content library / Research & data / Hypotheses, predictions, and competing explanations
Choose experiments that make competing explanations predict different results.
About 14 minutes, including the worked examples and checks.
Read a study and its figures · Lesson 1 of 6 · Continue in this path
On this page
- Concept: an explanation must make a prediction
- Pattern: separate a shared prediction from a distinguishing prediction
- Rule: make the conditional statement complete
- Application: follow the protein between compartments
- Use the 5P approach to choose the experiment
- Check your reasoning
- Continue learning
- Sources
Concept: an explanation must make a prediction
A culture releases more of protein P after a signal is added. That is an observation. “The signal causes the cells to make more P” is one explanation. “The signal causes the cells to release P they already stored” is another. Both explain the same increase in the surrounding medium. Repeating that measurement more precisely would establish the increase more convincingly, but it would not, by itself, tell the explanations apart.
A scientific hypothesis proposes a testable explanation. A prediction states what should be observed under particular conditions if that explanation and its supporting assumptions are correct. A model may combine several such claims into a simplified account of a system. Its usefulness depends partly on whether it gives observations a structure and generates predictions that can be challenged.1
On the MCAT, the useful question is often: What would the alternatives predict differently? That question connects a mechanism to the choice of an experiment. It also keeps the conclusion appropriately narrow. AAMC research-design skills include testable hypotheses, variable relationships, and limitations in research logic.2
Pattern: separate a shared prediction from a distinguishing prediction
Consider two deliberately simple models of the short-term protein response. Under the release model, the signal moves preexisting P from inside the cells to the medium without changing its combined amount. Under the synthesis model, the signal increases production of new P, some of which reaches the medium; for this model, assume there is no compensating increase in P loss during the measurement interval.
The models are explicit so that their predictions can be compared. They are not a complete list of everything a living cell might do. For example, changes in protein degradation or cell damage would require additional hypotheses and measurements.
| Measurement after the signal | Release model | Synthesis model |
|---|---|---|
| P in the medium | Increases | Increases |
| Combined P inside cells and in the medium | Remains approximately constant | Increases |
| Origin of extra P in the medium | Preexisting intracellular P | Newly made P |
The first row cannot discriminate: both models predict it. The second and third rows can. Measuring both compartments, or using an appropriate labeling strategy to distinguish newly synthesized from preexisting protein, addresses the disagreement. A sophisticated instrument is not automatically a better choice; the relevant measurement is the one that distinguishes the supplied predictions.
Rule: make the conditional statement complete
Write the prediction as: If this model is correct, and these measurement conditions hold, this result should follow. For the release model, the combined amount should remain approximately constant if both compartments are recovered, measured on a common calibrated scale, and compared over an interval with negligible net synthesis or loss. “Approximately” requires the precision of the assay, rather than an arbitrary demand for identical readings.
A result that conflicts with a prediction challenges the model or an assumption used to generate the prediction. Before choosing between those possibilities, check whether the intervention worked and whether the assay could detect the predicted difference. Conversely, a matching result supports a model without establishing that no other model could produce it. These are logical consequences of testing a conditional prediction, not reasons to avoid making conclusions.
An inhibitor illustrates the distinction. If blocking protein synthesis leaves the rapid increase in extracellular P intact, new synthesis is not necessary for that response under the tested conditions. But that inference depends on showing that synthesis was effectively blocked. If the inhibitor instead prevents release, it could have altered synthesis, cellular energy, viability, or secretion itself. The experiment becomes informative when those alternatives are addressed, rather than when the reagent is merely called a synthesis inhibitor.
Application: follow the protein between compartments
The following illustrative dataset represents matched cultures measured at the same endpoint. Each culture contains the same number of viable cells. The assay reports calibrated amounts of P in comparable units, and all medium and cellular material are collected.
| Condition | Inside cells (units) | Medium (units) | Combined (units) |
|---|---|---|---|
| No signal | 88 | 12 | 100 |
| Signal | 60 | 40 | 100 |
Start with the disagreement between the models. The medium gains 28 units while the cells lose 28 units, leaving the combined amount unchanged. Under the assumptions stated above, redistribution explains this pattern and the specified net-synthesis model does not. The claim “the signal increases the total amount of P” is unsupported because it substitutes a compartment measurement for a total.
The stronger conclusion is that these data favor redistribution over a net increase in P during this interval. They do not establish the molecular route of release. Nor do they exclude a more complicated model in which production and degradation increase by equal amounts. A pulse-labeling experiment could test that revised model because it asks about the age of the protein, a feature the amount measurements do not reveal.
What would challenge redistribution? Suppose both intracellular P and medium P increased, so the combined amount rose well beyond measurement error. Moving a fixed stock between two fully measured compartments could not explain that result on its own. A viable revision would need a source of additional P or evidence that the original measurement assumptions failed. Renaming the same unexplained pattern would not repair the model.
Use the 5P approach to choose the experiment
- Prioritize: identify whether the question asks for a prediction, a discriminating measurement, or a conclusion.
- Paraphrase: state each model in one causal sentence: “release existing P” versus “make additional P.”
- Prognose: predict the distinguishing pattern before reading the options: unchanged combined P versus increased combined P.
- Pick: choose the measurement or result that addresses that difference under the supplied conditions.
- Post-mortem: name the alternative explanation your selected experiment leaves unresolved.
This reasoning transfers to cell counts, behavioral responses, and chemical assays. A higher final cell count could reflect more divisions or fewer deaths; a stronger optical signal could reflect more product or a changed detector response. In each case, define the competing causes, identify their shared predictions, and measure a feature on which they disagree.
ACTIVE PRACTICE · COMPETING EXPLANATIONS
Sum the Compartments Before Choosing a Mechanism
The lesson’s experiment measures protein P inside matched cultures and in recovered medium. The release model redistributes a fixed stock; the specified synthesis model raises total P without compensatory increased degradation.
Inside: units; medium: units; combined: units.
Equal total at two times favors redistribution over the specifically defined net-synthesis model. It does not prove molecular trafficking or exclude balanced synthesis/degradation.
Check your reasoning
Choose an answer before opening its feedback. Each option explains the reasoning behind that choice.
Check 1
A compound increases the color intensity of an enzyme reaction. Model A proposes that it increases the amount of product. Model B proposes that it changes the product’s color without changing the amount formed. Which follow-up most directly separates these explanations?
A. Measure product quantity with an independently calibrated method unaffected by the compound.
Correct. Model A predicts more product; Model B predicts the same product quantity despite a stronger color signal. An independent measurement of the disputed quantity separates those predictions.
B. Measure the color at the original wavelength in additional reaction replicates.
Review this choice. More replicates improve the estimate of the color difference. Both models already predict that difference, so replication of the same readout does not distinguish its cause.
C. Repeat the reaction with twice as much substrate and measure its color.
Review this choice. Both product formation and the optical response can change the measured intensity. Without an independent quantity measurement or additional supplied predictions, this manipulation does not isolate the two explanations.
D. Use the original color measurement to estimate the amount of product formed.
Review this choice. That estimate assumes the relation between color and quantity has not changed—the very assumption Model B disputes. It cannot independently test that assumption.
The principle to carry forward
A useful test measures the point of disagreement without assuming the conclusion it is meant to test.
Check 2
Two matched cell populations begin with the same viable cell count. After treatment, one contains more viable cells. The proposed model states that treatment increases division frequency while leaving death frequency unchanged. Which result most directly challenges this model?
A. A second endpoint count again finds more viable treated cells.
Review this choice. The count repeats the observation to be explained. Both increased division and reduced death could produce it.
B. Lineage tracking finds unchanged division frequency and fewer deaths in treated cells.
Correct. The proposed model predicts increased division and unchanged death. Tracking instead identifies a different route to the higher count and contradicts both of those specific predictions.
C. Treated cells have a higher mean abundance of a cell-cycle-associated protein.
Review this choice. A protein association is not a direct measurement of division frequency. It may be compatible with the model but does not most directly challenge it.
D. Treatment produces a similar final count difference in a second culture batch.
Review this choice. Reproducibility of the final difference does not identify whether division, death, or both changed. The model requires a measurement of those processes.
The principle to carry forward
A final amount is often a balance between opposing processes. Measure the processes when the hypotheses disagree about which one changed.
Continue learning
Read a study and its figures
Lesson 1 of 6.
- Next in Read a study and its figures: Experimental Controls: What a Comparison Can Show
- Return to this path’s overview
See every step in Read a study and its figures
Read a study and its figures
Follow a question through its design, graph, normalization and statistical description.
- Hypotheses, predictions, and competing explanations · You are here
- Experimental Controls: What a Comparison Can Show
- Observational and experimental studies: what can we conclude?
- Reading graphs: axes, scales, and the comparison that matters
- Normalized data, percent change, and fold change
- Mean, median, and spread: describing a dataset
Related lessons
- Experimental Controls: What a Comparison Can Show
- Observational and experimental studies: what can we conclude?
- Normalized data, percent change, and fold change
Browse research & data · MCAT content library
MCAT preparation and roadmap · Aim High, Score Higher