
count kernel phenotypes on monohybrid and dihybrid corn ears, calculate expected counts, and use a chi-square test to decide whether the data fit a Mendelian ratio.
Phase 2: Investigate
Do Now. Your coin-toss partner got 26 heads in 40 tosses. Is the coin unfair, or is that just chance? Write your gut answer and one reason.
An ear from Pp × Pp has 381 purple and 119 yellow kernels. Expected is 375 and 125. Is the 3 : 1 model in trouble?
The problem with “close enough”. An ear from a Pp × Pp cross has 500 kernels: 381 purple and 119 yellow. The expected counts are 3/4 × 500 = 375 and 1/4 × 500 = 125. The data are off by 6 in each class. Is that chance, or evidence that the 3 : 1 model is wrong? A statistical test gives a principled answer.
Worked example 1: the chi-square statistic. For each category, square the difference between observed (O) and expected (E) counts and divide by the expected count, then add: χ2 = Σ(O − E)2/E. For our ear: (381 − 375)2/375 = 36/375 = 0.096, and (119 − 125)2/125 = 36/125 = 0.288. So χ2 = 0.096 + 0.288 = 0.384. Big differences give big values; perfect agreement gives zero.
Deciding. Compare χ2 with a critical value that depends on the degrees of freedom, one less than the number of categories. At the usual 0.05 significance level:
| Degrees of freedom | 1 | 2 | 3 |
|---|---|---|---|
| Critical value (0.05 level) | 3.84 | 5.99 | 7.81 |
If χ2 is below the critical value, the differences are small enough to be explained by chance, and we do not reject the model. Here 0.384 < 3.84, so the counts are consistent with 3 : 1. That is not proof the model is right; it means the data give no reason to doubt it.
Worked example 2: a dihybrid ear. Two genes: purple (P) or yellow (p), and smooth, starchy (S) or wrinkled, sweet (s). If a PpSs × PpSs cross follows independent assortment, the phenotypes come in 9 : 3 : 3 : 1. On an ear of 384 kernels, expected counts are 384 × 9/16 = 216 purple smooth, 72 purple wrinkled, 72 yellow smooth and 24 yellow wrinkled. Suppose you count 216, 79, 65 and 24 (illustrative). Then χ2 = 0 + 49/72 + 49/72 + 0 ≈ 1.36, with 3 degrees of freedom. Since 1.36 < 7.81, the data fit 9 : 3 : 3 : 1.
Worked example 3: when the model fails. Suppose another ear has 290 purple and 210 yellow. Then χ2 = (290 − 375)2/375 + (210 − 125)2/125 = 19.27 + 57.8 ≈ 77.1, far above 3.84. Reject 3 : 1 and ask what cross would explain it. A ratio near 1 : 1 would suggest a Pp × pp test cross, not Pp × Pp.
Counting well. Use actual counts, never percentages, in the formula. Mark a starting row with a pin so you do not count a row twice, and have a partner recount one row as a check.
Lab procedure: kernel counts on genetic corn ears. Question: do the kernel phenotypes on monohybrid and dihybrid ears fit the ratios Mendel’s model predicts?
| Ear | Phenotype | Observed (O) | Expected (E) | (O − E)2/E |
|---|---|---|---|---|
| Monohybrid | Purple | |||
| Monohybrid | Yellow | |||
| Dihybrid | Purple smooth | |||
| Dihybrid | Purple wrinkled | |||
| Dihybrid | Yellow smooth | |||
| Dihybrid | Yellow wrinkled |
No corn ears? Use the two illustrative row-by-row kernel tables your teacher provides, one monohybrid ear of 500 kernels in 10 rows and one dihybrid ear of 512 kernels in 8 rows. Total the columns yourself and run both tests; your answers should come out at 1.80 with 1 degree of freedom and 6.89 with 3. It earns full credit.
Reading: OpenStax Biology 2e, chapter 12, Mendel’s Experiments and Heredity (CC BY 4.0).
Use your counts and the critical-value table.
Finish your lab analysis.
Design a cross that would tell a PP purple plant from a Pp purple plant, and predict the kernel counts on an ear of 400 for each. Then identify two sources of counting error and whether each would push χ2 up or down.
The ratios hold, within chance. Next: where does new variation come from?
Exit ticket: in plain language for a seed-library volunteer, what does “χ2 = 0.384, below 3.84” mean?
How confident are you that you can count phenotypes, calculate expected counts and use chi-square to test a genetic ratio.
Materials: genetic corn ears from a biology supplier (one monohybrid and one dihybrid per group), push pins or sticky dots, tally counters, calculators or a free spreadsheet; illustrative row-by-row kernel tables for the no-ear route.
Timing: this lab spans two 50-minute periods: counting on day 1, chi-square tests and pooling on day 2.
Listen for: percentages put into the formula; “below the critical value proves the model”; the wrong degrees of freedom.
Support / stretch: give a chi-square calculation frame with one row worked. Stretch: compute how large an ear would need to be before a 381 : 119 proportion (76.2% purple) became a significant departure from 3 : 1.