The persistence of cumulative semantic interference during naming

• Cumulative semantic interference persists across time.
• Cumulative semantic interference dissipates across trials.
• Residual changes in activation do not account for cumulative semantic interference.
• Interference is persistent only in contexts which amplify semantic relationships.

Naming semantically related pictures (e.g., “goat” “cow” “mouse”) becomes increasingly slower when repeatedly naming from a semantic category even when several unrelated trials intervene (Howard, Nickels, Coltheart, & Cole-Virtue, 2006). The aim of this study was to test whether cumulative semantic interference is independent of time and unrelated trials between naming occurrences as predicted if interference is the result of learning reflected by persistent changes to semantic–lexical connection strength (Oppenheim, Dell, & Schwartz, 2010). Consistent with this account, changes in response stimulus intervals did not affect cumulative interference, allowing rejection of a temporary residual activation account of interference, suggesting that persistence of cumulative interference in this paradigm is similar to when exemplars are repeatedly named in the blocked-cyclic naming paradigm. However, cumulative interference disappeared when intervening unrelated trials increased (8–50) except when a short lag occurred (2) in the sequence. Critically, when a short lag occurred in the sequence, interference extended to new categories separated by long lags (8–14) which were not previously named. To account for results, modifications of learning models of naming should include a mechanism by which interference dissipates, and a mechanism which biases learning to create persistence in contexts where semantic relationships are amplified.