Historically, feeling and knowing have been viewed as distinct and sometimes competing psychological processes. Feeling involves emotions or values; pure knowledge is dispassionate. Three experiments examine an opposite view--that feeling (as evaluation) and knowing (as memory) are not two separate, distinct processes but are so fundamentally intertwined as to be psychologically indistinguishable. The logic of the three experiments is as follows: If knowing (memory) and feeling (affective evaluation) are two aspects of a single process, then changes in one should correspond to changes in the other. Past research has demonstrated reliable changes in serial position effects of free-recall with manipulations of presentation rate and recall delay: slower presentation rates enhance the primacy effect, delayed recall diminishes the recency effect. Experiment I attempted to determine if similar changes in the pattern of affective serial position effects could be produced by varying presentation rate and rating delay. Three lists of 12 nonsense words were presented to subjects at a two or four second rate. One group (N = 40) free-recalled the items, the other (N = 72) rated the items (7-pt. Good-Bad scale). Recall or rating was immediate or delayed 25 seconds. Primacy was obtained for both the recall and affect rating groups. Recency was obtained for recall and an "inferred" recency effect was obtained in rating. The delay diminished recall recency, and presentation rate had no effect on recall primacy. Neither delay nor presentation rate influenced rating serial position effects. Differences between rating and recall results were discussed. Experiment I suggested that serial position influences affective value. If knowing (memory) and feeling (evaluation) are two aspects of a unitary process, then affective value should influence serial position effects. Experiment II compared the serial position curve for nonwords rated "good" with that of nonwords rated "bad". It was expected that affective value would influence the shape of the serial position curve in ways similar to word frequency--i.e., a recency effect for free-recall of both "good" and "bad" nonwords, but a primacy effect with "good" nonwords only. Thirty-two students were given a single trial free-recall task using two lists of nonwords. One list consisted of 10 nonwords rated "good", and the other of 10 nonwords rated "bad". When recall of each list was analyzed separately, the expected results were obtained: Primacy with the "good" but not the "bad" list. Experiment III applied the approach used in the first two experiments to personality impression formation. Contrary to the affect-memory identity thesis, past research has shown that the evaluative impression formed of a hypothetical target person is not a simple function of the target's recallable traits. Several studies have reported impression primacy in conjunction with recall recency. In Experiment III, the standard Asch impression formation procedure was used: two lists of 12 trait words were read to each of 80 subjects who rated the likeability of the target person and recalled the list. Recall primacy and recency effects were manipulated as in Experiment I (varying presentation rate and delay) to test for similar order effects in impression formation. The correspondence between impression and recall order effects was strong. A slow presentation rate enhanced both recall primacy (with the second list) and impression primacy: initial trait words were better recalled and figured more prominently in the impression formed. A 25-second delay diminished both the recall recency effect and, among males, the impression recency effect: terminal traits were recalled less well and were less influential in determining the impression. Inconsistencies between the present results and those of previous research were discussed.