Priming in Recognition Memory for Categorized Lists - American ...

RECOGNITION PRIMING 567Table 2Average Reaction Times (in Milliseconds) and Accuracies forCritical Probes in Experiment las a Function of CategoryStatus and of List Membership of the PrimesCategory of primeSameMeanRTConditional accuracyUnconditional accuracyDifferentMeanRTConditional accuracyUnconditional accuracyExtra*740.77.90786.79.90List status of critical probeTargetIntra"696.79.90742.78.90Extra"825.77.87866.75.86LureIntra'817.75.86837.74.86Note. Only correct responses to both prime and critical probe are includedin the RT measure, and individual observations underlying eachsubject's mean are log-transformed, with values exceeding 2,000 ms notentering in the analysis. Means across subjects are based on antilogsof each individual's mean. Trimmed observations count as conditionalerrors, but are not part of the unconditional error measure.1 List status of prime.merit 1 that Neely et al. mentioned as a possible cause for theepisodic priming effect; namely, that lures were responded tomore slowly than targets. Episodically primed test items werethus always preceded by a faster response than test items thatwere not episodically primed. If that RT difference resultedfrom a difference in difficulty, this would be cause for concernbecause difficult items often exert an effect on subsequent easyitems in a test sequence (Kiger & Glass, 1981). However, giventhat response latencies to targets and lures can often be identical(e.g., Sternberg, 1966), it seems more reasonable to use accuracyas an index of difficulty of responding. Accuracy did notdiffer between lures and targets serving as primes (.863 and.870, respectively). Given the absence of any differences there,one should conclude that the episodic priming effect did notresult from a difference in difficulty in responding to theprimes.Overall, then, Experiment I showed that a facilitative semanticpriming effect can, under some conditions at least, obtainfor both targets and lures. The task now was to delineate theconditions under which inhibition is obtained for lure processing.In particular, the next step was to produce an inhibition oflure processing while controlling retention interval.Experiment 2This experiment was designed to test the type of prime processinghypothesis. If the type of processing were the criticaldeterminant of the effects of semantic priming, Experiment 2should show an inhibition for rejection of lures, because theprimes used here were similar to the ones used by Macht andO'Brien (1980). Whereas they used a sentence-verification taskDIFFERENT CATEGORY PRIMESAME CATEGORY PRIMEEXTRA INTRA EXTRALIST STATUS OF PRIMEFigure 1. Obtained reaction times for all conditions in Experiment 1.INTRA

RECOGNITION PRIMING 569Table 4Average Reaction Times (in Milliseconds) and Accuracies forCritical Probes in Experiment 2 as a Functionof the Category Status of the PrimeCategory of primeSameMeanRTConditional accuracyUnconditional accuracyDifferentMeanRTConditional accuracyUnconditional accuracyList status of criticalprobeTarget722.81.93771.80.90Lure948.73.86905.77.88Note. Only correct responses to both prime and critical item are includedin the RT measure, and individual observations underlying eachsubject's mean are log-transformed, with values exceeding 2,000 ms notentering in the analysis. Means across subjects are based on antilogsof each individual's mean. Trimmed observations count as conditionalerrors, but are not part of the unconditional error measure.within-subjects analysis. Overall conditional accuracy level wasat 80%. As in Experiment 1, unconditional error rates were thesame across both priming conditions for targets (.08) and lures(.13). No analyses were performed on this measure becausemeans did not differ within a response class. Table 4 presentsthe means for both conditional and unconditional accuracyscores.Response latency. During computation of each subject'smean, the same cutoff (2,000 ms) was used as in Experiment 1.This resulted in the loss of 2% of the observations. The 2X2ANOVA on RT means revealed that lures were responded tomore slowly (926.7 ms) than targets (746.2 ms), with F{ 1,15) =72.46, MS e = 7,195. Furthermore, the interaction betweenprime category and list status of the critical item was also foundto be significant, F{1, 15)= 1\.26,MS C = 3,010, whereas therewas no main effect of category of prime, F{\, 15) < 1. Figure 2and Table 4 display the RT means underlying this result. A posthoc exploration of the means using a Duncan's multiple rangetest indicated that, as would be predicted by the type of primeprocessing hypothesis, a same-category prime facilitated thesubsequent verification of a positive probe, whereas it inhibited(in comparison to a different-category prime) the rejection of asubsequent lure.DiscussionTo summarize, Experiments 1 and 2 both used the same listlength and almost identical retention intervals. Yet the resultswere quite different, depending on what type of prime was used.Consequently, retention interval could not have contributed tothe difference in outcomes. At the same time, the data reinforcethe presence of two discrepant outcomes in the recognitionpriming area. These patterns have now been shown to be replicableunder a variety of conditions. In particular, Experiment 2presents the first demonstration of a semantic priming inhibitionat longer retention intervals.Experiment 2 also provided the beginning of an explanationfor these discrepant outcomes, based on the type of processingrequired for the prime. The task used to achieve priming hasbeen clearly established as a critical determinant of what happensto processing of lures: A semantic and associative task resultedin inhibition (Experiment 2), whereas an episodic task(Experiment 1) resulted in facilitation. It is not clear yet, however,whether the priming task must be associative in nature, inaddition to being semantic, to produce an inhibition of lureprocessing. Experiments 3 and 4 were designed to investigatethe effects of an item-predicated yet semantic priming task.Experiments 3 and 4Experiments 3 and 4 observed priming as a function of anitem-predicated semantic task, by requiring a lexical-decisiontask to be performed on the prime. The lexical-decision taskcorresponds most closely to a single item episodic recognitiontask as it was used in Experiment 1. The only difference betweenExperiments 3 and 4, in turn, was the length of the study list.Experiment 3: MethodSubjects and arrangement of materials. Twenty-two subjects participatedin Experiment 3. Each category in the Categorized Word Poolwas randomly split into two sets (A and B) of 16 items each. Identity ofSets A and B was alternated for every second subject. Eight items wererandomly sampled from the respective Set A to form one category onthe study list, and eight items were selected from Set B to provide luresfor that category on the test list. Categories were randomly allocated totrials.For the test list, four nonstudied items from Set A were taken to functionas word (experimental) primes for that given category. Pseudowordprimes were obtained by changing the spelling of four members of SetB of that category such that they became pseudowords. A subject wouldnever see items from Set B and their pseudoword derivations togetherin the experiment. These pseudowords obeyed English spelling rulesand were pronounceable. They also maintained a resemblance to items1000 TARGETDIFFERENTCATEGORY PRIMESAMECATEGORY PRIMECRITICAL ITEMLUREFigure 2. Obtained reaction times for all conditions in Experiment 2.

570 STEPHAN LEWANDOWSKYTable 5Examples of a Study List and a Test List Usedin Experiments 3 and 4TOOLSJIGSAWBOLTSTest list pairword? HAMMERBOLTSword? WRENCHPSYCHOLOGYword? GEOGRAPHYASTRONOMYword ? AGRONOMYSCREWSStudy listTest listCategoryof primesamedifferentsamedifferentSCIENCESPSYCHOLOGYTOPOGRAPHYList status ofcritical probetargettargetlurelureNote. One example is given for each experimental condition. Fillerprobes and nonexperimental observations (i.e., pseudoword primes) areomitted. Right-hand columns specify levels of experimental variablesfor prime-critical probe pair on the left.from Set B of the category, which pilot work had shown to be necessaryin order to obtain a robust priming effect. For example, pseudowords forthe category PROFESSIONS were items such as MURCHENT, ARCHITOCT,JANETUR, and the like. Highly similar pseudowords have been used beforein priming research (cf. Antos, 1979;Lupker, 1984), and they presentan extension of other findings (Shulman & Davison, 1977; Smith etal., 1983) that priming effects are magnified if pseudowords are usedthat are more similar to real words. For the subsequent discussion, referto Table 5 for an example of the kind of stimulus arrangement used inExperiments 3 and 4.Recognition test items were distributed as follows: Half of all studyitems of a given category and half of the corresponding set of lures wererandomly designated to serve as fillerprobes. A fillerwas the firstmemberof each of the test triplets, similar to Experiment 2. Assign men I offiller items to position on the test list was random. Half of the remainingrecognition test items (two lures, two targets from each category) wereassigned to follow word (experimental) primes such that each experimentalcondition was represented once for each category. Critical responseswere observed only for those probes that followed word primes.As in Experiments 1 and 2, critical probes Owes or targets) were takeneither from the same category as the prime, or from the one differentcategory randomly yoked to it. Consequently, a given category item presentedas a prime was equally likely to be followed by a probe from thesame or from the different category. Similarly, because half of the criticalitems from a given category were targets and the others were lures,probability of a yes versus a no response following a word prime was alsoheld equal. Finally, nonexperimental primes (i.e., pseudowords) werefollowed by recognition test items whose properties were the same asthose following experimental primes. Recognition responses to probesfollowing pseudowords were of no interest. Primes and the probes followingthem were distributed throughout the test list in a random order.In summary, each category provided four critical probes (one in eachcondition) and also four word primes, yielding four observations percondition per trial per subject, or a total of 32 per subject for each condition.Procedure. Instructions emphasized both accuracy and latency.The study list was presented in the same fashion as in the precedingexperiments. Each triplet in the test sequence was shown as follows: Thefiller (a recognition test item) appeared on the screen and was, followingresponding, replaced 400 ms later by the prime (a lexical-decision item).The prime was printed in inverse video (i.e., dark green on a light greenbackground) with the prompt "word?" to the left of it. Upon performingthe yes/no lexical decision by using the usual response keys, theprime disappeared and was replaced 400 ms later by the critical probe,which was printed in normal video (i.e., bright green on a dark background).Again, following responding, the next triplet was presented400 ms later.Prior to onset of the experimental procedure, subjects received apractice trial consisting of 10 study items (the same across all subjects,taken from the Toronto Word Pool; Friendly, Franklin, Hoffman, & Rubin,1982), followed by a test sequence of 20 triplets similar in setup tothe experimental trials.Experiment 3: ResultsAccuracy. The 2X2 within-subjects analysis of conditionalaccuracy showed that there was a significant main effect of liststatus of the critical probe (targets: .80 vs. lures: .67), with F(l,21) = 40.7, MS e = .0094. In addition, the analysis uncovered aslight trend toward more accurate responding in the same categorycondition than in the different category condition, F{ 1,21) = 3.13,p

RECOGNITION PRIMING 571Table 6Average Reaction Times (in Milliseconds) and Accuracies forCritical Probes in Experiment 3 as a Functionof the Category Status of the PrimeCategory of primeSameMeanRTConditional accuracyUnconditional accuracyDifferentMeanRTConditional accuracyUnconditional accuracyList status of criticalprobeTarget884.78.93925.82.94Lure1256.65.781223.68.81Note. Only correct responses to both prime and critical probe are includedin the RT measure, and individual observations underlying eachsubject's mean are log-transformed, with values exceeding 3,500 ms notentering in the analysis. Means across subjects are based on antilogsof each individual's mean. Trimmed observations count as conditionalerrors, but are not part of the unconditional error measure.rately than targets, F(\, 19) = 16.11, MS C = .0161. Contrary toExperiment 3, however, there was no indication of an effect ofthe semantic priming manipulation. The grand mean of conditionalaccuracy was 77%, with a value of 0.83 for targets, and0.71 for lures, respectively. A similar difference between levelsof the list status variable was obtained for unconditional errors.Targets (.06 errors) and lures (.18) were responded to with thesame degree of accuracy regardless of priming condition. Table7 depicts the accuracy measures and also the RT means.Response latency. The trimming scheme was the same asthat used in Experiment 3 (3,500-ms cutoff) resulting in theloss of 1% of the observations. As in Experiment 3, critical lureswere responded to more slowly (1067.6 ms) than targets (839.3ms), with F{U 19) = 85.22, MS e = 12,233. The category of theprime had no effect with F{ 1, 19) = 2.39, MS* = 1,618. Instead,the interaction was again significant as shown by an F( 1, 19) =4.41, MS C = 3,932.Contrary to Experiment 3, the inhibition of lure processingwas the stronger effect this time, and it was significant using aDuncan's test. The facilitation of target verification, on theother hand, was reduced to a nonsignificant difference.DiscussionBoth Experiment 3 and Experiment 4 produced an interactionbetween the list status of the critical probe and the primingcondition. Target verification was facilitated by a prior lexicaldecision on a word from the same category, whereas processingof lures was inhibited. Although neither experiment on its ownwas characterized by a great statistical robustness, the patternof the interaction was replicated across both studies and bothretention intervals. The experiments showed that primingthrough a semantic task, even if that task is not associative, inhibitslure processing.This effect could not have been the result of the mere presenceof pseudowords in Experiments 3 and 4, because Experiment2 did not contain any pseudowords and yet showed thesame pattern of results. Similarly, an explanation for the resultsof Experiments 3 and 4 based on work by Balota and Chumbley(1984) can be ruled out. Balota and Chumbley suggested thatlexical decision may also involve an assessment of the familiarityof an item, similar in spirit to the Atkinson and Juola (1973,1974) recognition model. Thus, in Experiments 3 and 4, subjectscould have generalized such a familiarity-based strategyfrom the lexical-decision primes to the recognition probes.Such a strategy would of course predict inhibition of lure processing.But this explanation can be ruled out, given that Experiment2 involved a category membership judgment that, againaccording to Balota and Chumbley, does not involve a familiaritycheck.General DiscussionBefore one can discuss the implications of the results of thefour present experiments, several potential causes for concernand alternative explanations must be dealt with. First, it mustbe ensured that performance on the primes themselves did notcovary with performance on the critical items. Ideally, givenproper randomization, reaction time and accuracy of responsesto primes should not differ at all between conditions within anexperiment. Indeed, the observed variations between meanswere slight: For Experiments 1-4, the largest difference in reactiontime (computed with the appropriate trimming scheme)between primes for the different conditions was 20, 23, 16, and20 ms, respectively. The corresponding maximum differencesper experiment in accuracy were 1%, 5%, 3%, and 3%, respectively.Of the 32 F tests in ANOVAS that were performed on theprime data for all experiments, only one was found to be unexpectedlysignificant. (In addition, lure primes in Experiment Iwere responded to more slowly than target primes, but this isTable 7Average Reaction Times (in Milliseconds) and Accuracies forCritical Probes in Experiment 4 as a Functionof the Category Status of the PrimeCategory of primeSameMean RTConditional accuracyUnconditional accuracyDifferentMeanRTConditional accuracyUnconditional accuracyList status of criticalprobeTarget831.82.94847.83.96Lure1089.72.821046.70.83Note. Only correct responses to both prime and critical probe are includedin the RT measure, and individual observations underlying eachsubject's mean are log-transformed, with values exceeding 3,500 ms notentering in the analysis. Means across subjects are based on antilogsof the individual's mean. Trimmed observations count as conditionalerrors, but are not part of the unconditional error measure.

572 STEPHAN LEW\NDOWSKYwhat is often found in a recognition experiment, and is thereforenot unexpected). Consequently, differences in performanceon the primes could not have contributed to the results.Second, the overall performance level on the primes acrossexperiments must also be close to invariant in order to ensurethat primes were of roughly equal difficulty. Indeed, the grandmeans of accuracy did not vary much across studies, with valuesof 86%, 80%, 85%, and 86% for Experiments 1-4, respectively.Note how primes in Experiment 1 were responded towith precisely the same accuracy as those used in Experiment4. Yet, those two experiments uncovered widely different patternsof results.Third, in all experiments that showed an inhibition for therejection of lures, subjects had to switch from one task (categoryjudgment or lexical decision) to another (recognition)when responding to prime and critical item, respectively. Contraryto that, in Experiment 1, which obtained facilitation, subjectsperformed the same task on the prime and on the criticalitem. It is thus conceivable that this difference between experimentswas responsible for the different outcomes. However, thisappears unlikely, because responding to targets was facilitatedthrough semantic priming regardless of whether the primingtask was the same as the recognition task. It is implausible toassume that task switching was responsible for drastically alteringresponses to critical lures, given that it had no effect on targetverification. Furthermore, I have conducted several otherunpublished experiments that required task switching betweenprime and critical probe, but that nevertheless demonstrated afacilitation for lures. Taken together, these two points shouldallay any fears that the present results were an artifact of thetask switching required in Experiments 2-4 but not in Experiment1.The fourth and last issue concerns another proceduraldifference between Experiment 1 on one hand and Experiments2-4 on the other hand. None of the primes used in the last threeexperiments had been presented to subjects at study. Contraryto that, half the primes used in Experiment 1 had been studied.The reason for this procedural difference between experimentslies in the essence of the task dimension that was manipulated.An episodic task can only be performed after a prior studyphase, whereas a semantic task can be (and usually is) performedon novel items that subjects have not seen previously inthe same experiment. However, it is, in principle, possible toperform a semantic judgment on an item that has been studiedpreviously (e.g., Neely & Durgunoglu, 1985). The present setof experiments cannot serve to elucidate that special case ofepisodically studied primes that require semantic responses.Thus, for the summary below, it must be kept in mind thatprimes requiring episodic and semantic responses always involvedstudied and novel items, respectively.Summary of ResultsThe results from all four experiments can then be summarizedas follows: If a purely semantic decision is required for theprime, subsequent processing of negative recognition probes isinhibited. This is the case regardless of whether the semanticjudgment is one that is predominantly item predicated or onethat requires associative information. If, on the other hand, aprime is used that requires an episodic judgment, processing oflures is facilitated. This facilitation is increased if the prime hadbeen present on the study list.The effect of primes on target probes, on the other hand, doesnot appear to vary as a function of the nature of the prime processing.That is, verification of targets is facilitated after an episodicas well as after a semantic priming task. Again, for a primerequiring an episodic judgment, that facilitation is greater whenthe prime had been studied.This pattern is clearly inconsistent with the retention intervalhypothesis of Neely et al. (1983). Inhibition of lure processinghas now been obtained at retention intervals ranging from 15 s(Macht & O'Brien, 1980) through approximately 60 s (Experiment4) to 120 s (Experiments 2 and 3). Conversely, facilitationhas been obtained at retention intervals ranging from 80 s (Experiment1) through 150 s (Neely et al., 1983) and several minutes(Johns, 1985) up to 24 hr (Taylor & Juola, 1974). We nowknow that the type of prime processing, not retention intervalor list length, is the critical predictor of outcomes in priming ofrecognition memory. The retention interval hypothesis of Neelyet al. should thus be discarded unless future research can proveits usefulness. Beyond rejection of that hypothesis, however,how can the different effects of different priming tasks be explained?From a purely heuristic point of view, the episodic/semanticdimension is the most useful one to capture the pattern of results.That is, we can now predict an outcome on the basis ofknowledge of the type of processing that is required for theprime. A semantic task yields inhibition, whereas an episodictask results in facilitation. Beyond its predictive capabilities, theepisodic/semantic dichotomy can only provide a rather vagueframework for an explanation for the present data. Accordingto that vague framework, the data may be interpreted as showingthat priming may sometimes be restricted to episodic memoryonly, in which case the consequences of priming are differentfrom the situation when all category items (studied or nonstudied)are primed. If episodic memory (i.e., the list) is primedonly (through a recognition probe functioning as prime), subsequentepisodic processing is facilitated. On the other hand, ifthe entire category is primed (through a prime requiring a semanticjudgment), subsequent rejection of a lure is inhibitedbecause the familiarity of all items in a category has beenraised.If this reasoning is put into the more detailed terms of anAtkinson and Juola type model, from which all research in recognitionpriming appears to depart, an account for the findingsin the present paper can be proposed: (a) A prime that requiresa semantic judgment raises the level of activation of all items ina category, studied or nonstudied. Familiarity-based responsesto targets are thus facilitated, whereas these responses are inhibitedfor lures. This mechanism explains all inhibitory effects,(b) A prime that requires an episodic judgment, on the otherhand, activates the studied list only. Consequently, targets arefacilitated because the familiarity of old items from that categoryhas been raised. To explain the effect on lures, one mustassume that priming of the list also speeds up the search process(as suggested by Neely et al., 1983) for subsequent probes andthat this has the same quantitative effect on lures that the risein familiarity has on targets, (c) When the prime is not from the

same category, there is nevertheless some residual priming ofthe list, provided the prime requires an episodic judgment andis itself a target. This explains purely episodic priming for targetsin Experiment 1. To account for the effect on lures, it mustagain be assumed that residual list priming also speeds up subsequentsearch.This account is highly speculative and clearly post hoc, butcomponents of it have been suggested previously. Macht andO'Brien (1980) derived their predictions by using point a fromabove. Neely et al. (1983) assumed a speeding of search to underliefacilitation for lures (see points b and c). What the previouspoints do, then, is to integrate elements from the previousliterature and to specify the conditions under which each ofthese processes is relevant. Because those conditions (semanticvs. episodic tasks required on the primes) are denned independentlyand can be known a priori, the account can serve to derivepredictions for other priming studies-ConclusionThe present experiments have demonstrated that primingcan facilitate or inhibit processing of lures even when list lengthand retention interval are controlled. The critical variable thatdetermines whether facilitation or inhibition arises is the typeofjudgment that must be performed on the prime. If a semanticjudgment is required (regardless of what type of information isinvolved) lure processing is inhibited. If the prime requires anepisodic judgment, on the other hand, subsequent rejection oflures is facilitated.These data can be explained if it is assumed that the list canbe primed selectively. The idea of selective list priming is consistentwith the view that episodic and semantic memory may constitutetwo different memory systems (Tulving, 1983). The ideais also consistent with the work of Wickens and his associates(Wickens, Moody, & Dow, 1981; Wickens, Moody, & Vidulich,1985) who also demonstrated that lists of varying lengths arestored in memory as a well-defined and closely interlinked entity.ReferencesAnderson, J. R. (1976). Language, memory, and thought. Hillsdale, NJ:Erlbaum.Anderson, J. R. (1983a). The architecture of cognition. Cambridge, MA:Harvard University Press.Anderson, J. R. (1983b). A spreading activation theory of memory.Journal of Verbal Learning and Verbal Behavior, 22, 26-295.Antos, S. J. (1979). Processing facilitation in a lexical decision task.Journal of Experimental Psychology: Human Perception and Performance,5, 527-545.Atkinson, R. C, Herrmann, D. J., & Wescourt, K. T. (1974). Searchprocesses in recognition memory. In R. L. Solso (Ed.), Theories incognitive psychology; The Loyola symposium (pp. 101-146). Potomac,MD: Erlbaum.Atkinson, R. C, & Juola, J. F. (1973). Factors influencing speed andaccuracy of word recognition. In S. Kornblum (Ed.), Attention andperformance IV(pp. 583-612). New York: Academic Press.Atkinson, R. C, & Juola, J. F. (1974). Search and decision processes inrecognition memory. In D. H. Krantz, R. C. Atkinson, R. D. Luce,& P. Suppes (Eds.), Contemporary developments in mathematicalpsychology (pp. 243-293). San Francisco: Freeman.RECOGNITION PRIMING 573Balota, D. A., & Chumbley, J. I. (1984). Are lexical decisions a goodmeasure of lexical access? The role of word frequency in the neglecteddecision stage. Journal of Experimental Psychology: Human Perceptionand Performance, 10, 340-357.Bennet, R. W. (1975). Proactive interference in short-term memory:Fundamental forgetting processes. Journal of Verbal Learning andVerbal Behavior, 14, 123-144.Blaxton, T. A., & Neely, J. H. (1983). Inhibition from semantically relatedprimes: Evidence of a category-specific inhibition. Memory &Cognition, 11,500-510.Collins, A. M., & Loftus, E. F. (1975). A spreading-activation theory ofsemantic processing. Psychological Review, 82,407-428.deGroot, A. M. B., Thomassen, A. J. W. M, & Hudson, P. T. W. (1982).Associative facilitation of word recognition as measured from a neutralprime. Memory & Cognition, 10, 358-370.Friendly, M., Franklin, P. E., Hoffmann, D., & Rubin, D. C. (1982). TheToronto Word Pool: Norms for imagery, concreteness, orthographicvariables, and grammatical usage for 1080 words. Behavior ResearchMethods & Instrumentation, 14, 375-399.Gorfein, D. S., & Jacobson, D. E. (1973). Memory search in a Brown-Peterson short-term memory paradigm. Journal of ExperimentalPsychology, 99, 82-87.Henik, A., Friedrich, F. J., & Kellogg, W. A. (1983). The dependenceof semantic relatedness effects upon prime processing. Memory &Cognition, 11, 366-373.Herrmann, D. J., & Harwood, J. R. (1980). More evidence for the existenceof separate semantic and episodic stores in long-term memory.Journal of Experimental Psychology: Human Learning and Memory,6, 467-478.Irwin, D. I., & Lupker, S. J. (1983). Semantic priming of pictures andwords: A levels of processing approach. Journal of Verbal Learningand Verbal Behavior, 22, 45-60.Johns, E. E. (1985). Effects of list organization on item recognition.Journal of Experimental Psychology: Learning, Memory, and Cognition,11, 605-620,Kiger, J. I., & Glass, A. L. (1981). Context effects in sentence verification.Journal of Experimental Psychology: Human Perception andPerformance, 7, 688-700.Lewandowsky, S. (1982). An examination of various control conditionsfor priming research in recognition memory. Unpublished manuscript,University of Toronto.Lorch, R. F. (1982). Priming and search processes in semantic memory:A test of three models of spreading activation. Journal of VerbalLearning and Verbal Behavior, 21,468-492.Lupker, S. J. (1984). Semantic priming without association: A secondlook. Journal of Verbal Learning and Verbal Behavior, 23, 709-733.Macht, M. L., & O'Brien, E. J. (1980). Familiarity-based responding initem recognition: Evidence for the role of spreading activation. Journalof Experimental Psychology: Human Learning and Memory, 6,301-318.Macht, M. L., & Spear, N. E. (1977). Priming effects in episodic memory.Journal of Experimental Psychology; Human Learning andMemory, i, 733-741.McKoon, G., Ratcliff, R., & Dell, G. (1985). The role of semantic informationin episodic retrieval. Journal of Experimental Psychology:Learning, Memory, and Cognition, 11, 742-751.Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation in recognizingpairs of words: Evidence of a dependence between retrieval operations.Journal of Experimental Psychology, 90, 227-234.Monsell, S. (1978). Recency, immediate recognition memory, and reactiontime. Cognitive Psychology, 10, 465-501.Murdock, B. B., Jr. (1974). Human memory: Theory and data. Hillsdale,NJ: Erlbaum.Murdock, B. B., Jr. (1976). Item and order information in short-term

574 STEPHAN LEWANDOWSKYserial memory. Journal of Experimental Psychology: General, 105,191-216.Neely, J. H., & Durgunoglu, A. (1985). Dissociative episodic and semanticpriming effects in episodic recognition and lexical decisiontasks. Journal of Memory and Language, 24,466-489.Neely, J. H., Schmidt, S. R., & Roediger, H. L., III. (1983). Inhibitionfrom related primes in recognition memory. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 9,196-211.Posner, M. I., & Snyder, C R. R. (1975). Attention and cognitive control.In R. L. Solso (Ed.), Information processing and cognition: TheLoyola symposium. Hillsdale, NJ: Erlbaum.Ratcliff, R., & McKoon, G. (1981a). Automatic and strategic primingin recognition. Journal of Verbal Learning and Verbal Behavior, 20,204-215.Ratcliff, R., & McKoon, G. (1981 b). Does activation really spread? PsychologicalReview. 88, 454-462.Shulman, H. G., & Davison, T. C. B. (1977). Control properties of semanticcoding in a lexical decision task. Journal of Verbal Learningand Verbal Behavior, 16.91-98.Smith, M. C. (1979). Contextual facilitation in a letter search task dependson how the prime is processed. Journal of Experimental Psychology:Human Perception and Performance, 5, 239-251.Smith, M. C, Theodor, L., & Franklin, P. E. (1983). The relationshipbetween contextual facilitation and depth of processing. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 9, 697-712.Steinberg, S. (1966). High-speed scanning in human memory. Science,153, 652-654.Taylor, G. A., & Juola, J. F. (1974). Priming effects on recognition performance.Bulletin ofthePsychonomic Society, 3, 277-279.Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W.Donaldson (Eds.), Organization of memory (pp. 381-403). NewYork: Academic Press.Tulving, E. (1983). Elements of episodic memory. New \brk: OxfordUniversity Press.Wickens, D. D., Moody, M. J., & Dow, R. (1981). The nature and timingof the retrieval processes and interference effects. Journal of ExperimentalPsychology: General, 110, 1-20.Wickens, D. D., Moody, M. J., & Vidulich, M. (1985). Retrieval timeas a function of memory set size, type of probes, and interference inrecognition memory. Journal of Experimental Psychology: Learning,Memory, and Cognition, 11, 154-164.Received June 24,1985Revision received September 30,1985 •Low Publication Prices for APA Members and AffiliatesKeeping You Up-to-DateAll APA members (Fellows, Members, and Associates) receive — as part of their annual dues — subscriptions tothe American Psychologist, the APA Monitor, and Psychology Today.High School Teacher and Student Affiliates receive subscriptions to the APA Monitor and Psychology Today, andthey can subscribe to the American Psychologist at a significantly reduced rate.In addition, all members and affiliates are eligible for savings of up to 50% on other APA journals, as well assignificant discounts on subscriptions from cooperating societies and publishers (e.g.. the British PsychologicalSociety, the American Sociological Association, and Human Sciences Press).Essential ResourcesAPA members and affiliates receive special rates for purchases of APA books, including the PublicationManual of the APA, the Master Lectures, and APA's Guide to Research Support.Other Benefits of MembershipMembership in APA also provides eligibility for low-cost insurance plans covering life: medical and incomeprotection; hospital indemnity; accident and travel; Keogh retirement; office overhead; and student/school.professional, and liability.For more information, write to American Psychological Association,Membership Records. 1200 Seventeenth Street NW. Washington. DC 20036