WEAVERϩϩ has no backward links in its form-production network and yet is able to explain the lexical and mixed error biases and the mixed distractor latency effect. This refutes the claim of B. Rapp and M. Goldrick (2000) that these findings specifically support production-internal feedback. Whether their restricted interaction account model can also provide a unified account of the error biases and latency effect remains to be shown.There were two major points in my comment (Roelofs, 2004) on the article by Rapp and Goldrick (2000). My first major point was that the lexical and mixed error biases and the mixed distractor latency effect do not specifically support production-internal feedback, contrary to what Rapp and Goldrick (2000) maintained. I showed that a model with comprehension-based rather than production-internal feedback, namely WEAVERϩϩ (e.g., Levelt, Roelofs, & Meyer, 1999;Roelofs, 1992Roelofs, , 1997Roelofs, , 2003aRoelofs, , 2003bRoelofs & Hagoort, 2002), accounts for the error and latency findings. My second major point was that extant productioninternal feedback accounts of the error biases and latency effects are incompatible.In their reply, Rapp and Goldrick (2004) agreed that the empirical evidence leaves open whether the feedback is comprehension based or production internal: "We find no clear evidence in support of the claim that feedback in spoken word production is unambiguously a component of the comprehension process. However, it is important to be clear that neither is there clear evidence that it is not" (p. 578). Moreover, they were silent on the issue of whether the production-internal feedback in their restricted interaction account (RIA) model serves speech comprehension. It is important to be clear, however, that regardless of the possible comprehension purposes of the backward links in RIA, the account of the findings by Rapp and Goldrick (2000) is in terms of feedback that occurs in a production network, which differs from my account in terms of two strictly feedforward networks, one for production and the other for comprehension, as implemented in WEAVERϩϩ. I refer to Roelofs (2003b) for a review of the evidence for distinct word-form networks for production and comprehension. On my account, feedback arises because the feedforward production network activates the feedforward comprehension network. A secondary issue is whether activation of the comprehension network by the production network happens automatically, as suggested by the new evidence of Humphreys (2002) and Nooteboom (in press) referred to by Rapp and Goldrick (2004), or whether it is under voluntary control, as I assumed on the basis of the available evidence (Roelofs, 2004). Either way, the feedback does not happen production-network internally, which is the critical point.The second major point of my comment (Roelofs, 2004) was that extant production-internal feedback accounts of the mixed error bias and the mixed distractor latency effect are incompatible. Rapp and Goldrick (2004) disputed this claim. Accordin...