RREPAER os00 On the Status of Inhibitory Mechanisms in Cognitio: 1 Aman Pog Amen Memory Retrieval as a Model Case Michael C. Anderson University of California, Los Angeles Barbara A. Spellman University of Texas at Austin “Theories of cognition frequently assume the existence of inhibitory mechanisms that deactivate mental representations. Justifying this assumption i dificult because cognitive effects thought to reflect inhibition can often be explained without recourse to inhibitory processes. This article ad- {resses the uncertain status of cognitive inhibitory mectanisms, focusing on ther fonction in mem ‘ory retrieval, On the basis ofa novel orm of forgetting ceported here i shown that classical ‘associative theories of interference are insuficient a accounts of forpeting ad that inhibitory pro- ‘cesses must be at work. It is argued that inibiory processes are uzed to resolve computational problems of selection coraman to memory retrieval and seletive attention an that retrievals bes regarded as conceptually focused selective attention ‘Advances in both brain and computational sciences have fu- led an explosion of theories positing inhibitory mechanismsin 4 variety of cognitive Functions. Theories of intact cognition ‘emphasize the importance ofthese activation-reducing mecha- nisms to basic capacities, including visual selective attention (Duncan & Humphreys, 1989; Keele & Neill, 1978; Neill & Westberry, 1987; Neumann & DeSchepper, 1992; Tipper, 1985; Tipper, MacQueen, & Brehaut, 1988), language comprehen sion and production (Gernsbacher, 1990, 1991; Gernsbacher & Faust, 1991; Simpson & Kang, 1994), and retrieval from epi sodic and semantic memory (Anderson, Bjork, & Bjork, 1994; Bjork, 1989; Blaxton & Neely, 1983; Carr & Dagenbach, 1990), aswell as to some higher order functions such as analogical rea- soning (Holyoak & Thagard, 1989; Spellman & Holyoak, 1993) ‘Michael C. Anderson, Department of Psihology, University of Cal fornia, Los Angeles; Barbara A. Spellman, Department of Psychology, University of Texasat Austin, “This work was done while the authors were graduate students together atthe University of California, Los Angeles it constitutes one chapter jn Michael C. Anderson's dsertation, which appearson University Mi- ‘es and word are shown in Appendis A. Relatednes of category pairs ‘sa8 manipulated in the study phase of the experiment by varying the ‘combinations of categories studied by participant; Table | ilustrates the combinations ofeatgoris studied by various participants in the experiment ‘Second. we manipulate the retival-pratice status of an exemplar on fou level: Rb, Ro, Nep-Dissmilar, and Nrp-Similar Asin the ‘Anderson et ai, (1994) procedure. participants performed retrieval practice on half of he studied eategories; thus, fora given participant, two ofthe four studied categories were practiced (Rp) categories, and two were unpracticed (Nep) categories. For each Rp category, atic ‘pants practiced three exemplars (Rp+ items) and did not practice the Femaining tree (Rp~ items). For example a5illutrated in Table 1 for Participant I, some participants practiced the exemplars RED-BLOOD, RED-FIRE, and RED-SUNBURN from the Rp category RED, making those exemplars Rp* items: the remaining three unpracticed exem= lars fom hat category RED-TOMATO, RED-APFLE, and RED-CHERRY, then became Rp items. For each Nepeategory, three ofthe sx exem- pars were similar 10 other items the experiment (Nrp Similar items) land three were dissimilar to other items in the experiment (Nrp- Dis Similar items). For example, again asshown for Participant | ip Table |, some participants studied the Following exemplars ofthe Nrp category FOOD: STRAWBERBY, RADISH, KETCHUP, CRACKERS, PEAS, 2nd BREAD. The first three exemplars would be Nep-Similar items because they ae also semantically red things, bu the second three exemplars would be [Nep-Dissrier lems because they are not also red things, Note that, fer Participant 1, the Nep category FOO was in the elated condition, sothe Nrp Sila items wer similar to other items actualy stadied by ‘the participant (ie. some ofthe foods were als red). Foran Nrp cate ory inthe uncelated condition (ea, LOUD for Participant 1), NED Similar items (LAWNMOWER, SANDBLASTER, and COMPRESSOR) were Similar to other items we constucte (i, the tools FILE, PLIERS, and SCREWORIVER), but those other items were not studied by the Participant, As illustrated by the preceding examples, and as shown in Table 1 the Rp# exemplars inthis experiment were always dissimilar items, and the Rp~ items were similar items. Note that because participants practiced only dissimilar items, the sets of Rp+ and Nep-Dissimilar items were identical, as were the sets of Rp~ and NrpSimiar Kem, alter consideration ofall of, our counterbalancing measures. We re ‘corded the percentage of items corel rersfed inthe final category ‘ved recall test.(ON THE STATUS OF INHIBITORY MECHANISMS 15 Table f Mlustration of Learning~Practice Combinations for Half ofthe Participants in Experiment 1 Learning Booklet 1 Learning Booklet 2 Learning Booklet 3 Category Panicipant Panicipant? Participant 3 Paricipant 4 Participant $ Participant Res Blood Blood Blood Blood Fire Fire Fire Sunbura Sunburn Sunburn Apple ‘Apple ‘Apple Cherry Cherry Cherry Tomato Tomato Tomato Food Breas Bread Bread Bread Crackers Crackers Crackers Crackers Peas Peas eas Peas Ketchup Ketchup Ketchup Ketchup Radish Radish Rais Radish Strawbery Strawberry Strawberry Strawbery Fy Kite Kite Kite Glider Glider Glider Frisbee Frisbee Frisbee Butterfly Buttery Buttery Eale Eagle Eagle idyoug Ladybug Ladybux Animal Sheep Sheep Giraffe Grate Hamster Hamster Hamster Wap Wasp Wasp Bat Ba Bet Pigeon Pigeon Pigeon Lous ‘Thunder ‘Thunder ‘Thunder ‘Thunder yell Yel Yell Yel Trafic Trafic ‘Trathe Trafic Lawntnower Lawnmower Lawamower Lawnmower Sandbiastes Sandblaster Sandblaster Sandblasier Compressor Compressor Compressor Compressor Too! File Fite File File Piers Pers Pliers Pliers Seromdrver Screwdriver Screwdriver Screwdriver Sackharmer Jackhammer Jackhammer Jackhammer Dnt Dai Dail Dri! Chainsaw Chainsaw Chainsaw Chainsaw Nove. Related pars of categories are vertically adjacent. Each participant studied items down the coum Practiced items appear in boldface. Within eac categors: the fist thre items ate dismilar items; the Tas thre are similar tems. Six additonal participants would be necessary to show the complete design: for them, the similar items would be exchanged between related categories. Procedure ‘We collected data from the participants individually or in groups of| po 10, They were old that they would be participating i an exper- ‘ment on memory and reasoning and thatthe experiment would have Several pars. The experiment was conducted in four phases learning ‘hase, a rerival-prctce pase e distractor phase, and a surprise cat egory-cued recall phase. Learning Phase Panicipants were gien learning booklet and were od that (a)each ‘age inthe booklet contained a eategory-example par and they were to try to remember te category examples as bes they could (b) they would be given 5 sto study each catepory-example pair and they were ‘supposed to spend al oF his time relating the example tits category (c)atthe endaf5s.a voice ona tape recording would instruct them to turn he page; and (his Sequence was tobe repeated untill pages in ‘the learning booklet had been studied, Retrieval-Practice Phase ‘When participants completed the learning phase, the experimenter collected learning booklets and distributed practice booklets, Partici- [pants were told that (a) each page i the new booklet contained one ‘ofthe category names that they had seen inthe previous phase of the ‘experiment and the frst two letters of one ofthe members ofthat cate- ‘gory (b) they would be given 10 sto try to remember the example and to write it down on the page, (c) they were to retrieve something that they had already seen rather than simply figure out any word that ft, and (d} some ofthe examples might be tested more than once, in which case they were supposed do the same thing each time. Asin the eamn- ing phase, tape-recorded voice instructed them when 0 turn pages. Distractor Phase ‘Afr te retrieval practice phase had been completed, booklets were ‘collected, and participants were given an unrelated reasoning ask that {0k 20 min16 MICHAEL C. ANDERSON AND BARBARA A. SPELLMAN Final Test Phase Pantsipants were given recaifbookfts anc were colt that (a) ae topot each page, tere would be the name of one ofthe categories stud- ied previously (b) they would be ven 30 tory to recall all examples ofthat category that they had seen a anytime in the experiment; and (hater 30 the experimenter would insirvet them to turn ta the next ‘page. When participants completed this ask, they were debriefed and dismissed Materials Categories ‘We constructed 10 categories, 4 filler and 6 experimental, The 6 eX perimental categories consisted of three pars of relate categories (Appendix A provides the complete list of tems forall three of these pais of categories). Each category of aie contaned three items that ‘were uniquely members of that particular category (i, dissimilar items}. Across the relied pair of extesories, however, there were six items that semantically were members of both categories (i, similar stems) For example, the categories RED and FooD formed one pair oF related categories. RED included tree disimilar items such as BLOOD; "FOOD included tee dissimilae items such ab CRACKERS. In addition, there were six items such a TOMATO and STRAMBERKY that were similar items (three for each category). Note that simile tems could be cate- orzed under either category ofthe related pair; dissimilar items could not. The filler categories mountains, cities, countries, and compos exs—consised of si items each, Category members were selected according 10 the following con- straints Firs, we quired tat exemplars be unambiguous memes of ‘their categories. Fr example, we rejected he item DRESS as an exer lar ofthe RED category because, although dresses are often red they are not necessarily re. no are dresses red by default. We selected the item BLO0D as a member ofthe RED category, however, because most ‘eaple assume blood tobe red and red is strongly associated with the ‘wont 8.009, Sas, se excluded exerplarsif they had sang astoc tons with other exemplars that had alreadybeen selected, to minimize Daricipants' ability tous nterexempla retrieval strategies during the final test phase Third to ensure that each exempt’ two-letter cue in the retrieval-practice task uniquely designated that exemplar, we did notallow two words inthe experiment to egin with the sae first (0 eters, Finally to avid interference from extra-experimental tems, We 4 not allow any chosen categary member to have the same frst 160 Jeers asa familiar unchosen category member For example, we ex ‘ded the item Boas an ANIMAL Because participants might intrude Learning Booklets Learning booklets consisted of 48 pages, cach containing the name of| ‘category and an exemplar ofthat atezory 0 its right (eg. RED '8L000), centered on the page. Each booklet containedallsixexemplas ‘rom the four ier categories, aswel as six exemplars—chree disimtae and three simila—from each of four of ou six experimental rategoies. ‘The four experimental eategories alvays consisted of one ofthe pir of elated categories and one eategory fFom each ofthe remaining {0 ‘elated pairs, implementing our related So unrelated conditions, re specively (sce Table 1). We ordered these categories ad exemplars in the learning booklet in such a way a8 to aod the formation of both withincategory and between categories associations. Categories were prohibited from appearing twice in a 10%, and related eatepories were prohibited from occurring adjacently. In addition, we generally pre vented unrelated estegories fom appearing next to each other more han oDee. To cont) For seria) postion elects, we ensured shat the average study position foreach category (and, therefor, forthe unce- lated’ vs. related and the to-bepracticed vs. notto-bepraciced ‘ategories) vas approximately equal These restrictions applied to the experimental categories: filler catepores were used liberally to enforce the constraints for the experimental categories. The fst 3 and ast 3 ‘ages ofthe learning booklets contained filler category items to contcot For primacy and recency effets. Each of our categories appeared equaly often in the elated and un- related conditions In addition, aye critica similar exemplars appeared ‘equally often in both related categories (e.g, TOMATO was studied hall ‘ofthe time undee FOOD and half under RED). The former but no latter ‘manipulation illustrated in Table | Retrieval-Practice Booklets Retsieval practice booklets consisted of 1 paps, each containing the name ofa studied category and the first two eters ofa studied member ofthat category followed by a slid blank ine (eg, RED 8), ‘centered onthe page, We held the length ofthe blank line constant so as to sive no cus for word length. ‘We constructed retrievalpractice booklets as follows. Each booklet tested participant on sx categories: four filer nd two critical exper ‘mental eatepories, The two experimental categories always included one related and one unrelated category. From thes two categories, we se lected only the dissimilar items forthe retrieval practice booklet. Table 1 shows practiced items for etch patiipant in boldface. We con- structed the practice booklets by inserting each eiical exemplar into the booklet three times in an expanding sequence. The frst and second practices occurred with an average of 30 intervening items; the second fnd thtd practives oceurte with an average of 67 intervening items. To reduce epportunites for iterexacation among exemplars pro hibited members ofthe seme category from appearing adjacently inthe booklet, and, in general, we did not allow items to be practiced adj ‘cently moce than once. To minimize order eects, we equated the aver ‘abe practic position for items in the two practiced categories. These resitions applied tothe experimental categories agai, filler eatepory items were used liberally 10 enforce the previous constaiats. Pier items always occupied the first and lst thee practice positions inthe booklet ach of our experimental categories was practiced and unpractied ‘cqually often (compare Participants {and 2in Table 1, who learned the same four categories but practiced 1w0 complementa'y eateries) In Addition, to ensure that the particular way in which we inserted exem Blarsinto the practice booklets had no impact on our results, we rested ‘wo different random practice orders foreach set of practiced items Recall Booklets Recall booklets consisted of four pages, each containing te name of| 1 learned experimental category atthe top. We consructed the final Fecal booklets as follows The fst page always contained the name of a ategry in the related condition, the second page contained the name of category ithe unrelated condition, the third page contained the name of the other related category, and the fourth page contained the ame ofthe other unrelated category. For half ofthe participants, the practiced categories were recalled fst and ls; forthe other hall ofthe Participants, the practiced categories were recalled second and tied (as result of rexreval-practice counterbalancing). Thos, related catego ries were never rraled adjacently, Acros al participants, the average ‘output position ofthe practiced and unpracticed categories was the same. Different ext booklets were constr ueted to correspond with each Ikarning booklet so that participants recalled itemsnly fom catepores they hag studied. There were two eiferent recall orders foreach earn ing bookletON THE STATUS OF INHIBITORY MECHANISMS 1 Results and Discussion All analyses were performed with learning booklet and prac- tice counterbalancing as between-subjects conditions. Retrieval Practice We counted whether participants successfully retrieved the correct item during each retrieval practice trial for the exper ‘mental (not fille) eategories. Because participants practiced six ‘experimental items (three from one of the categories in the lated condition and three from one of the categories in the un- related condition )each three times, there were 18 relevant prac: tice trials. The retreval-practice success rate fr items from the ‘category in the unrelated condition (69%) did not difersig- nificantly from the retreval-practice success rate for items from. the category in the related condition (73%), F(1, 42) = 1.04, Final Recall Note, for purposes of comparing recall percentages, that in this design Rp+ and Nrp-Dissimilar items were identical, as ‘were Rp and Nrp-Similar items (see Table 2). We report the ‘Table 2 Experiments 1,2, and 3: Mean Percentage of Words Recalled ‘as a Function of Caregory Relatedness and Resrieval-Practce Status of ems Retrieval practic sats Practiced category ‘Unpacticed category (te) (F000) Caegory Rpt Rpm Nips. Nop telatedness (61000)_(TOMATO) (STRAWBERRY) _ (CRACKERS) Experiment | (W = 48) Unrelated 69 Pn “4 Related 74 2 36 Practiced category Unpracticed category (onsen) (sours) Category Rpt Rp- Nwws ND reltedness (EMERALD) (LETTUCE) (MUSHROOM) _ (CHICKEN) Experiment 2(N = 54) Unrelated 2 a [2] a Related 6 36 |37) 3 [Experiment 3a (N= 27) Unrelated @ 36 Related 8 rn Note. Examples of tems nach condition ae shown in parentheses; they corespond tothe examples in Figures 2 and 4. Unrelated cate fore were leamed witout chit pared category in the leaming ‘ookettelated eategories were leamed with ther paired category in the learning boollet. The comparisons for detecting cross-eategory Innibiton for these experiments are shown in boxes. RP-+ = practiced items; RP— = unpracticed items; NepS = similar items, NipD