Mollusk shells from Peruvian archaeological middens provide brief (< 5 yr per shell) records of past marine conditions. Marine radiocarbon age, R, is recorded in shell carbonate at the time of precipitation. R varies with changes in upwelling: when radiocarbon-depleted sub-thermocline water wells up, R is large; increased contribution from radiocarbon-enriched surface water (due to seasonal cycles or an El Niño event) reduces R. Are molluscan records of R a useful proxy for Peruvian upwelling? If so, does R from archaeological shells reveal mid-Holocene upwelling changes that constrain the Holocene history of El Niño-Southern Oscillation (ENSO)? Profiles of R along ontogeny from early 20th century Argopecten purpuratus (bay scallop) shells and mid-Holocene A. purpuratus, Mesodesma donacium (surf clam), and Trachycardium procerum (cockle) shells from eight coastal Peru locations show that R varies by up to 530 ± 200 ¹⁴C yr within individual shells. El Niño events are easily detectable in post-1950s shell carbonate due to increased radiocarbon contrast between sub- and super-thermocline water from “bomb carbon,” but R differences between El Niño and La Niña shells from the early 20th century are subtle. Decreasing precision in older shells due to ¹⁴C decay makes detecting El Niño events in the archaeological past using radiocarbon very difficult. Because of intrashell radiocarbon variation, caution is prudent when using marine material for chronometry in variable upwelling environments. Based on modeling, mollusks that grow seasonally rather than year-round can skew long-term average (> 1 yr) R reconstructions by nearly 200 ¹⁴C yr toward R of the preferred growth season. Coldloving M. donacium, for example, records older marine reservoir ages on average than A. purpuratus in the same water, because A. purpuratus grows in both warm and cold conditions. Comparisons of R between species with opposite seasonal growth habits can compound this effect. Because of intrashell R variation, seasonal growth biases, and measurement uncertainties, a change in R due to past ENSO changes would have to be hundreds of ¹⁴C yr or greater to be identifiable. Thus far, clear evidence for such a Holocene change in R has not been seen.