Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: J Exp Psychol Learn Mem Cogn. 2015 Aug 17;42(3):339–350. doi: 10.1037/xlm0000136

Category Cued Recall Evokes a Generate-Recognize Retrieval Process

R Reed Hunt 1, Rebekah E Smith 1, Jeffrey P Toth 2
PMCID: PMC4757519  NIHMSID: NIHMS687380  PMID: 26280355

Abstract

The experiments reported here were designed to replicate and extend McCabe, Roediger, and Karpicke’s (2011) finding that retrieval in category cued recall involves both controlled and automatic processes. The extension entailed identifying whether distinctive encoding affected one or both of these two processes. The first experiment successfully replicated McCabe et al., but the second, which added a critical baseline condition, produced data inconsistent with a two independent process model of recall. The third experiment provided evidence that retrieval in category cued recall reflects a generate-recognize strategy, with the effect of distinctive processing being localized to recognition. Overall, the data suggest that category cued recall evokes a generate-recognize retrieval strategy and that the sub-processes underlying this strategy can be dissociated as a function of distinctive versus relational encoding processes.

What did you do yesterday? What medications do you currently take? Who did you see at the conference? Questions such as these are common queries of memory requiring recall of multiple responses to a single cue. This everyday demand on memory is investigated in the laboratory with tests of “free recall” which, like the examples above, is initiated by general cues but is “free” in the sense that output order is not constrained. A major focus of research on free recall has been the characterization of the underlying retrieval processes. To date, that effort has produced substantial if not unanimous agreement that successful free recall results from a cue directed search process. Although differing in their underlying assumptions, most prominent theories of recall assume that a controlled search process is sufficient to describe retrieval in free recall (e.g., Atkinson & Juola, 1974; Gillund & Shiffrin, 1984; Humphreys, Bain & Pike, 1989; Jacoby, 1991; Mandler, 1980; Tulving, 1985), and assume that, in contrast to recognition, automatic familiarity plays no role in free recall because in free recall no item is presented to generate a familiarity signal (Quamme, Yonelinas, Kroll, Suave, & Knight, 2004; Wixted & Squire, 2010).

Against this context, McCabe, Roediger, and Karpicke (2011) reported data that appear to contradict the prevailing theoretical view. McCabe et al. (2011) were motivated by prior research hinting at automatic influences on free recall, much as familiarity is assumed to exert an automatic effect in recognition. To investigate this possibility, McCabe et al. used Jacoby's (1991) process dissociation procedure (PDP), which assumes that retrieval is affected by both controlled and automatic influences of prior experience, which independently contribute to performance. Application of the PDP to recognition and cued recall has produced impressive support for the independence of controlled and automatic process underlying retrieval (for reviews see Aggleton & Brown, 2006; Yonelinas, 2002; Yonelinas & Jacoby, 2012). But prior to McCabe et al.’s report no published research had applied the PDP to free recall. The lack of research can be attributed to the widely held belief that free recall is a pure measure of recollection and, as noted above, does not include automatic influences such as familiarity (Aggleton & Brown, 1999; Quamme, et al. 2004; Tsivilis et al., 2008; Yonelinas et al., 2002). To test this belief, McCabe et al. used the PDP (e.g., Jacoby, Toth, & Yonelinas, 1993) to determine if the obtained process estimates revealed an influence of automatic processing. They also manipulated full versus divided attention at encoding to determine if the process estimates were consistent with the independent-process assumptions underlying the procedure. As described in more detail below, the pattern of process estimates obtained by McCabe et al. mimicked prior research examining the effect of attention on controlled and automatic influences on memory, leading to the conclusion that the data “demonstrate a contribution of automatic processing in free recall” (McCabe et al., 2011, p 402).

McCabe et al.'s findings challenge the widely endorsed view that automatic processes play no role in free recall. Their findings also have the potential to inform interpretation of other established findings in areas such as aging and amnesia. For example, many studies have reported that amnesic patients’ primary deficit is in recall while recognition is spared in various degrees due to the automatic influence of familiarity (e.g., Dorfman, Kihlstrom, Cork, & Misiaszek, 1995; Hirst, Johnson, et al., 1986; Hirst, Johnson, Phelps, & Volpe, 1988; Huppert & Piercy, 1976; Huppert & Piercy, 1977). Similar results have been found for the effects of aging on memory (e.g., Craik & McDowd, 1987; Dankert & Craik, 2013). Jacoby and his colleagues (e.g., 1999; Cermak, Verfaellie, Butler, & Jacoby, 1993) have analyzed both amnesia and aging effects and have concluded that the principal deficit is in recollection while automatic processing is preserved. This analysis fits the data well, assuming that recognition reflects both recollection and familiarity while recall relies on recollection alone. However as McCabe et al. note, free and cued recall performance of amnesic participants often is well above 0 (e.g., Graf, Squire, & Mandler, 1984), which could be an indication of automatic influences on free recall.

Given the importance of McCabe et al.’s (2011) conclusion for theories of recall as well as for application of those theories, we conducted a series of experiments with two primary goals. The first was a straightforward attempt to replicate the main findings of McCabe et al. The second was to extend the analysis to the well-documented effect of distinctive processing (see Hunt, 2006, 2012, 2013). The reason for choosing distinctive processing as an analytic target is that it is assumed to affect recollection, not automatic processing, in free recall. This assumption is based on the function of distinctive processing in providing highly diagnostic information that allows precise discrimination between targets and competitors, a function not served by automatic influences (Dobbins, Kroll, Yonelinas, & Liu, 1998). However, the only attempts to analyze distinctive processing in terms of recollection and automatic processing have been on tests of recognition (Dobbins et al., 1998; Hunt, 2003). These prior analyses showed effects of distinctive processing on both recollection and automatic processing. Given these results, it seems possible that the effect of distinctive processing in recall is also mediated through automatic processing.

To preview our studies, we were able to successfully replicate McCabe et al.'s key findings in Experiment 1. However, in Experiment 2, the extension of their paradigm to include distinctive processing and the inclusion of a key baseline condition raised serious questions about McCabe et al.’s conclusion that free recall involves two independent processes. Experiment 3 shows that a dependent-process generate-recognize strategy provides a better account of performance in free recall and localizes the effects of distinctive processing to the recognition component of this strategy.

Experiment 1

Our first experiment was a replication of McCabe et al., which used an inclusion/exclusion variant of the process dissociation procedure (Jacoby et al., 1993) applied to category cued recall. The PDP assumes that the effects of prior experience on current performance are conveyed either through recollection of prior episodic details or through automatic influences of the prior experience. Further, the theory assumes that these processes operate independently and that their engagement results from direct access to the prior experience. Testing this assumption requires gaining estimates of recollection and automatic processing in order to determine the effect of variables on those estimates. For McCabe et al. the variable under investigation was full versus divided attention at study. The effect of attention on the theoretical estimates is crucial for establishing the validity of the theory for a given situation because prior research has shown that dividing attention at encoding impairs subsequent recollection but has little or no influence on more automatic forms of memory (Eich, 1984; Koriat & Feuerstein, 1976; Parkin, Reid, & Russo, 1990).

In McCabe et al. (2011) participants studied a categorized list of words under full or divided attention followed by two different tests. One of the tests, inclusion, required recall of studied words in the presence of the category labels. The second test, exclusion, required generation of category instances to category labels, half of which had been used in study, with explicit instructions not to produce any studied items. According to the theory, the probability of correct inclusion performance is represented by the independent influence of recollection (R) and automatic access (A) of the studied word:

I = R + A (1 − R)

On the other hand, successful exclusion performance requires that the studied item be recollected lest the automatic influence of that experience evoke the item. Thus the theory specifies exclusion errors result from automatic access in the absence of an opposing influence from recollection:

E = A (1 − R)

With the data from the inclusion and exclusion tests, McCabe et al. (2011) used the equations to determine the effect of dividing attention on estimates of recollection and automatic influences.

The estimates obtained by McCabe et al. (2011) were consistent with predictions from Jacoby’s (1991) dual process model. Dividing attention at study reduced the estimate of recollection but had no effect on the estimate of automatic processing. Our first experiment attempted to replicate this result using the same procedure. In addition, our design included a manipulation of orienting tasks at study in order to test the notion that distinctive processing effects in recall result from recollection, not automatic processes. Distinctive processing allows for discrimination among otherwise similar items, thereby enhancing accuracy of memory performance (Hunt, 2006). In the context of a categorized list, pleasantness rating of each item draws attention to differences among the categorically similar items because pleasant rating requires detection of meaning and the meaning of each instance is different. In contrast, a task that orients attention to the category similarity discourages individuation of the items and will produce less accurate memory (e.g., Hunt, 2003). Previous research on the effects of pleasantness ratings has consistently found it facilitates recollection but in all cases the comparison condition has been a non-semantic encoding task (e.g., Cohn, Moscovitch, & Davidson, 2010; d’Ydewalle & van Damme, 2007; Horton, Wilson, Vonk, Kirby, & Nielsen, 2005; Newell & Andrews, 2004; Richardson-Klavehn, Gardiner, & Ramponi, 2002; Toth, Reingold, & Jacoby, 1994). Both of our encoding tasks are semantic and although we expected to replicate previous results showing that enhanced performance on the inclusion and exclusion tests following pleasantness rating is reflected in estimates of recollection, previous results were based on semantic-nonsemantic contrasts. Our contrast is between two semantic tasks, where one theoretically encourages distinctive processing and the other does not.

Method

Participants and design

Participants were 160 undergraduate volunteers whose participation partially fulfilled a course requirement for introductory psychology. They were assigned randomly to one of four conditions. The conditions were defined by the orthogonal combination of attention and encoding tasks, with attention being either full or divided at study and the tasks being either category judgment or pleasantness rating.

Materials

A pool of 120 words was selected from the following four categories used by McCabe et al (2011): body parts, four-legged animals, sports, and articles of clothing. These are large categories from which many items can be generated, an important criterion given that the exclusion test requires generation of non-studied items. The words used were taken from the Van Overschelde, Rawson, and Dunlosky (2004) norms. The 30 most frequent responses for a given category, excluding obsolete and identical exemplars, were selected. These 30 items were further separated into two sub-lists, using every other frequency-ranked word. The sub-lists were then combined over categories to create two separate 60-word study lists containing 15 items from each of the 4 categories. Presentation of the lists was counterbalanced such that an equal number of subjects saw each list. The items within the lists were blocked by category. A test booklet was used to record responses to both types of test. Each page of the booklet had a category label at the top with 15 answer spaces underneath. Both the inclusion and exclusion test included two of the four studied categories and the categories were counterbalanced across subjects such that each category appeared equally often in each test.

Procedure

The experiment was conducted with 1-2 participants in each session. Upon entering the laboratory, they were seated at individual work stations and informed consent for participation was obtained. Instructions informed the subjects that they would see a list of words presented one at a time on the monitor. In the category judgment condition, they were told to determine the category membership of each word and press the number key corresponding to the four categories listed below the word. The category labels were numbered 1-4 and accompanied each word. Subjects in the pleasantness rating condition were instructed to read each word and rate its subjective pleasantness on a scale from 1-4. As with category judgments, the numerical scale accompanied each word. The instructions included screen shot examples of a word and the rating scale.

Subjects in the divided attention conditions also were required to engage in a digit detection task during the study presentation (cf. Craik, 1982). A random series of the digits 0-9 were played auditorially and continuously throughout the study session. Subjects were instructed to press a counter each time they heard the digits 4 or 9. They were then fitted with headphones and adjusted the sound as needed. Two minutes of practice on digit detection was followed by an additional two minutes of practice on digit detection along with the encoding task. Immediately after this the study list was presented. Each study item remained on the monitor for 3 seconds, a presentation rate previously determined to be adequate for the encoding tasks.

A two minute distractor task followed the study list. A letter of the alphabet appeared on the monitor and subjects were instructed to press the key corresponding to the letter that followed the presented letter alphabetically as fast as possible. The inclusion test followed immediately after the distractor task. Participants were given a test booklet containing two pages with a category label at the top of each and received the following instructions taken from McCabe et al. (2011):

Earlier you saw 15 words from the category ‘x’ [e.g., body parts]. I want you to try to recall all 15 of the words on the answer sheet. If you cannot recall all 15 words (and most people cannot), I want you to guess which words you believe might have been on the list (they should be body parts) to fill in all 15 spaces. Don’t worry if you misspell words, but please do not use any slang words.

On completion of the inclusion test, another two page booklet was distributed along with instructions for the exclusion test again taken directly from McCabe et al.

On this next test I want you to write 15 words for the category ‘x’ [e.g., sports], but none of them can be sports that you saw earlier in the experiment. So, you must write 15 new sports that were not presented. All the words should still fit the category. Don’t worry if you misspell words, but please do not use any slang words.

Subjects were told that they could have much time as needed to complete each test.

Results

Prior to statistical analyses, the data from 6 full-attention participants (3.8% of total) were dropped because their exclusion performance was at ceiling (zero errors), a result that precludes application of the PDP as will be described in the Discussion section for this experiment. Four of these participants were from the pleasantness rating condition and two were from the category judgment condition.

Inclusion/Exclusion Performance

The results from the inclusion and exclusion tests are shown in the top portion of Figure 1 as a function of encoding task and attention condition. On the inclusion test, pleasantness ratings of the study items led to higher recall than category judgment, F(1, 150) = 5.45, p = .02, ηp2 = .03. Recall was reduced by dividing attention, F(1, 150) = 12.63, p = .001, ηp2 = .08. The interaction between encoding task and attention condition was not reliable, F < 1, p = .55. As can be seen in Figure 1, the proportion of exclusion errors was substantially lower following pleasantness ratings than category judgments, F(1, 150) = 39.33, p < .001, ηp2 = .21. Exclusion errors were reliably higher following divided as compared to full attention, F(1, 150) = 6.09, p = .01, ηp2 = .04. The interaction of the two variables did not reliably affect exclusion errors, F < 1, p = .82.

Figure 1.

Figure 1

Open in a new tab

The proportion of studied items produced and estimates of recollection and automaticity as a function of test type, attention, and study task in Experiment 1. Error bars are standard errors.

Estimates

Theoretical estimates of recollection and automatic processing were obtained by entering the inclusion and exclusion data into the simultaneous equations representing Jacoby’s (1991) dual process theory. The obtained estimates are shown in lower portion of Figure 1 as a function of encoding task and attention condition. Examining first the critical issue of the effect of attention on performance, Figure 1 shows that estimates of recollection were considerably lower under divided than full attention, but attention seemed to have little impact on estimates of automatic processing. Analyses bore out these conclusions. Estimates of recollection were reliably higher in full than in divided attention, F(1, 150) = 16.65, p < .001, ηp2 = .10, but no reliable difference was found in the automatic estimates as a function of attention, F(1,50) = 1.35, p = .25. Encoding task also reliably affected recollection with pleasantness ratings yielding higher estimates than category judgments, F(1, 150) = 34.23, p < .001, ηp2 = .19. Interestingly, encoding task exerted an equally impressive effect on the estimate of automatic processing with category judgments leading to much higher estimates than pleasantness ratings, F(1, 150) = 31.44, p < .001, ηp2 = .17. The interaction between encoding task and attention condition was not reliable for estimates of either recollection or automatic processing, Fs < 1, ps > .57.

Discussion

As expected, performance on the inclusion and exclusion tests was affected both by encoding task and divided attention. Correct recall was poorer and intrusion errors more frequent when attention was divided. Pleasantness rating enhanced correct recall and reduced false recall relative to category judgment, replicating previous studies of distinctive processing (e.g., Hunt, Smith, & Dunlap., 2011; Hunt & Rawson, 2011). More importantly, the theoretical estimates obtained from the inclusion and exclusion tests replicated the findings of McCabe et al. (2011) in showing that divided attention affected estimates of recollection but left estimates of automatic processing invariant. This result supports McCabe et al.’s conclusion that automatic processes play a role in free recall and provides further encouragement to adopt independent dual process theory as the description of retrieval processes in this type of test.

The theoretical estimates for the effects of the encoding tasks provide interesting new information. Pleasantness ratings led to higher estimates of recollection than category judgments. Assuming that distinctive processing is processing difference in the context of similarity (Hunt, 2012), pleasantness ratings encourages distinctive processing, and the recollection estimates obtained here converge with results from studies of recognition memory (Dobbins, et al., 1998; Hunt, 2003). An unexpected finding was the inverse relation that emerged between encoding task and type of estimate. That is, pleasantness ratings produced higher recollection estimates but lower estimates of automatic processing than category judgments. This inverse relationship has two possible interpretations. Perhaps category judgment increases automatic processing in this paradigm where list words are to be included or excluded based on the category label cue. Alternatively, the inverse relationship may signal violation of assumptions underlying Jacoby’s dual process theory (Jacoby, 1998).

In their investigation of word-fragment cued recall, Curran and Hintzman (1995) found that increased study time increased estimated recollection but decreased automatic estimates. They argued that the inverse pattern resulted from a lack of independence between the two processes. In the ensuing debate, several limitations of the independence model were identified (for reviews, see Jacoby, 1998; Jacoby et al., 1997; Yonelinas & Jacoby, 2012). For one, ceiling performance on the exclusion test (no errors) yields an automatic estimate of zero and results in an underestimation of automatic influences when aggregated over subjects. The underestimation of automatic influences can produce an inverse pattern of estimates because a variable that enhances recollection is also likely to reduce exclusion errors. Another important limitation is that direct retrieval instructions are necessary to satisfy the independence assumption. Instructions that emphasize a post-retrieval monitoring strategy, such as "examine each of your exclusion responses carefully to be sure they were not in the list," will often yield data violating independence (Jacoby, 1998). We were careful to follow both of these prescriptions in the present experiment, dropping subjects with ceiling performance on the exclusion test and using direct retrieval instructions. The question is: Does the inverse relationship between the estimates and encoding task accurately reflect the processes underlying performance or is it an indication that the independence dual process model does not apply to this type of recall test? The second experiment addresses this question.

Experiment 2

In his user’s guide to the PDP, Jacoby (1998) listed a number of circumstances that can lead to violations of the theory’s assumptions along with ways to determine if such violations have occurred. Base rate data – that is, the production of unstudied target items – are a critical source of diagnostic information. Unfortunately, neither our first experiment nor those of McCabe et al. (2011) included base rate conditions. Base rates can index the criteria participants used for responding in the inclusion and exclusion tests, where the assumption of equal criteria in the two tests is critical for use of the theoretical equations. Unequal base rates may signal violation of the independence assumption. For example, a lower base rate in the exclusion compared to the inclusion test would be consistent with a generate/recognize strategy. This is so because base rate items, none of which were previously studied, sometimes would be falsely recognized in exclusion as having been studied, leading to a lower exclusion base rate (Jacoby, 1998). Base rates also are necessary to accurately estimate the true influence of automatic processing (Jacoby et al., 1993). The production of instances to a category label at test results from both the influence of the prior study experience and one's accumulated experience with the category. The latter is reflected in the base rate data. Thus, subtracting base rates from the automatic estimate provides an estimate of the true influence of study. Based on the common assumption that item exposure during study elevates automatic access, one expects to find estimates of automatic influence to be higher than the base rates but in the absence of the base rate comparison, estimates of automatic influence of study items alone are uninterpretable.

The second experiment replicated the design of the first experiment with the addition of base rate (unstudied) categories in both the inclusion and exclusion tests. The goal of the second experiment was to establish the reliability of the estimates for the encoding tasks from the first experiment, as well as to provide a basis for diagnosing the inverse relationship that was found between those estimates.

Method

Participants and design

A total of 372 undergraduate volunteers received partial credit toward a course requirement for their participation. Participants were randomly assigned in groups of 2-5 to one of the four conditions defined by the orthogonal combination of attention and encoding task. The design was the same as in the first experiment with the addition of unstudied categories to the inclusion and exclusion test booklets.

Materials and procedure

A pool of 120 words was established by selecting 15 instances from each of 8 categories. Four of the categories were those used in the first experiment and the other four were countries, birds, musical instruments, and states. The 15 instances were selected to represent the range of frequencies for their category in Van Overschelde et al.’s norms. Frequencies were approximately equal across the categories. Two separate study lists were created from the 15 items from four categories. At test, four category labels were provided for the inclusion test and four for the exclusion test. Two of the labels in each test corresponded to studied categories and two to non-studied categories. The categories were counterbalanced across the studied/non-studied lists and inclusion/exclusion tests such that an approximately equal number of subjects saw each category in each condition.

The procedure was the same as in the first experiment with one addition. Instructions for the inclusion and exclusion tests added the proviso to produce the first 15 instances of a category that came to mind if there was no memory of studying the category.

Results

Prior to analysis, the data from 17 participants (4.6% of total) were dropped. For 12 of these (9 from pleasantness and 3 from category judgment) exclusion was at ceiling (no errors). The other 5 (2 pleasantness and 3 category judgment) reversed the inclusion and exclusion tests as evident in their post-experiment comments to the experimenter.

Inclusion/Exclusion performance

The data for the inclusion and exclusion tests are shown in the top portion of Figure 2. As is obvious in the figure, dividing attention reduced correct recall in inclusion, F(1,351) = 30.57, p < .001, ηp2 = .08, and increased errors on the exclusion test, F(1, 351) = 6.27, p = .02, ηp2 = .02. Pleasantness ratings led to higher correct recall than category judgments on the inclusion test, F(1, 351) = 7.23, p = .007, ηp2 = .02, and to fewer errors than category judgments on the exclusion test, F(1, 351) = 60.38, p < .001, ηp2 = .15. The interaction between attention and encoding task was not reliable for either inclusion or exclusion, Fs < 1, ps > .30.

Figure 2.

Figure 2

Open in a new tab

The proportion of studied items produced and estimates of recollection and automaticity as a function of test type, attention, and study task in Experiment 2. Error bars are standard errors.

Estimates

The estimates for recollection and automatic processing are shown in the lower portion of Figure 2 as function of attention and encoding task. The estimate for recollection was lower in the divided relative to the full attention condition, F(1, 351) = 26.32, p < .001, ηp2 = .08, but dividing attention had no reliable effect on the estimate of automatic processing, F(1, 351) = 1.46, p = .23. Pleasantness ratings yielded higher estimates of recollection than category judgments, F(1, 351) = 54.94, p <.001, ηp2 = .13. Automatic estimates, however, were higher following category judgments than pleasantness ratings, F( 1, 351) = 57.12, p < .001, ηp2 = .14. Thus the inverse relationship between encoding task and theoretical estimates found in Experiment 1 was replicated in the second experiment.

In order to determine the implications of the inverse relationship for the use of Jacoby’s (1991) process dissociation theory, the base rate data were examined as prescribed by Jacoby (1998) for detecting violations of his theory. Base rates were analyzed as a function of study task, attention, and test instructions. Neither attention, F < 1, p = .60, nor study task, F < 1, p = .40, reliably affected base rate production of target items. Test instructions (inclusion/exclusion) did reliably affect base rates, F(1, 353) = 9.82, p = .002, ηp2 = .03. Exclusion base rates (M = .35, SE = .01) were lower than inclusion base rates (M = .37, SE = .01).

Base rates also were compared to the estimates of automaticity. For this analysis, the data were collapsed over attention because attention did not reliably affect the estimates or base rates. The base rates used in the analysis were the average inclusion/exclusion rates. Analysis of these data yielded an interaction between measure (base rate, automatic estimate) and study task, F(1, 353) = 51.94, p < .001, ηp2 = .13, which was investigated with separate analyses performed for each encoding task. For category judgment, the automatic estimate (M = .36, SE = .01) did not differ reliably from the base rate (M = .35, SE = .05), F< 1, p = .77. For pleasantness rating, the automatic estimate (M = .26, SE = .01) was reliably below the base rate (M = .37, SE = .005), F(1, 170) = 75.77, p < .001, ηp2 = .31.

Discussion

The second experiment successfully replicated the inclusion/exclusion performance and pattern of theoretical estimates from the first experiment. The data from the two experiments along with McCabe et al. (2011) provide support for the validity of the dual process model in the form of invariance of automatic estimates following full and divided attention. Critically, however, these results also are consistent with a dependent process, generate-recognize model of performance. Based on research using category generation as an implicit memory test, dividing attention has been shown to have little effect on generation but leads to considerably lower performance on explicit recognition memory tests (e.g., Mulligan & Stone, 1999). Thus, the data on divided attention alone do not unambiguously favor the independence model. Moreover our data contain a warning in the form of the inverse relationship between the estimates for the two encoding tasks, suggesting that the independence model might not be appropriate for this recall test. To resolve the ambiguity, we turn to the analysis of the base rate data from Experiment 2. Those results send two clear signals that the assumption of independence between recollection and automatic processing is not met in this situation.

First, the base rate of target item output in the exclusion test was lower than that in the inclusion test. This outcome conflicts with the assumption that the criteria for producing old items on the two tests are the same (Jacoby, 1998). The difference in base rates is consistent with a generate- recognize theory of retrieval whereby false recognition of baseline items in the exclusion test can account for the difference in base rates. A generate-recognize strategy violates independence because recognition is dependent upon prior generation. The second clear indication that the independence model has been violated is the fact that estimates of automatic processing were below the base rate for the pleasantness rating condition. Outside of a violation of the independence assumption, such negative “priming” is not easily explained (see Curran & Hintzman, 1995; Jacoby, 1998). Because we tried to avoid violating the assumptions by using direct retrieval instructions and by eliminating subjects who performed at ceiling on exclusion, the data from Experiment 2 suggest performance in free recall is not the result of independent recollection and automatic processes as suggested by McCabe et al. (2011). The final experiment tests the viability of a generate-recognize model as a description of the effects of the encoding tasks applied to categorized material.

Experiment 3

The version of generate-recognize theory we adopt is that proposed by Jacoby and Hollingshead (1990). This model incorporates the general assumption of all versions of generate-recognize theory in that retrieval entails two processes. The initial stage is generation of candidate responses in response to cues. The generation stage is then usually followed by a recognition process which monitors generated items for accuracy. Jacoby and Hollingshead add to these standard assumptions the caveat that not all items are submitted to the monitoring process. Items generated with sufficient fluency are output without undergoing the recognition process. They also add the important qualification that generation is affected by particular prior events, such as the study experience.

Jacoby and Hollingshead prescribe a paradigm to test the validity of the generate-recognize model. The paradigm includes three independent test conditions following the study experience. One of the conditions is essentially an implicit memory test that requests generation of items to cues without reference to the study list. A second condition is the same generation test with the additional instruction to recognize any list items that are generated. These two conditions provide diagnostic information to determine if recall data, provided by the third condition, conform to a generate-recognize strategy. Clear predictions concerning the pattern of output from the three conditions follow from the assumptions made by Jacoby and Hollingshead’s model. Among those are that recall of studied items should never exceed studied item production in the generation condition because generation is a requisite stage of recall. Recall may be less than generation because of the possibility of mistakenly identifying a generated study item as non-studied. For this same reason, recall should either equal or exceed old item production in the generate-recognize condition because some generated items in recall may not be subjected to recognition (due to fluency) whereas all items in the generate-recognize condition are subjected to a recognition check (as per task instructions).

In addition to the general criteria for validating the model, predictions also can be made about the effects of distinctive processing on performance in the three test conditions. As discussed earlier, distinctive processing is encoding of differences among list items in the context of encoded similarity among the items. Pleasantness rating of categorized lists is operationally distinctive processing in that the categorical relationships are assumed to be processed spontaneously and pleasant judgments are based on meaning of each item, all of which are different. We expect recall following pleasantness judgments to exceed recall following category judgments, replicating the first two experiments. Is the effect of distinctive processing on generation, recognition, or both? Given the categorical structure of the lists, encoding of categorical relationships is critical for generation of study list items to the category label at test. Our assumption is that both study conditions foster encoding of categorical information and hence we predict that the study-task manipulation will have no effect on indices of generation. Measures of recognition, however, are expected to be higher following distinctive processing than following category judgment. This prediction is based on findings from previous research using recognition memory tests in which comparisons between conditions comparable to those used here showed an advantage to distinctive processing (e.g., Hunt, 2003, 2013). Thus on the generate-recognize model, the advantage of distinctive processing over processing of similarity lies in the recognition of generated items.

Method

Participants and design

The participants were 300 undergraduate volunteers who received course credit for their participation. Participants were randomly assigned to one of 6 conditions defined by the orthogonal combination of study task and type of test. Half of all subjects performed pleasantness ratings and half category judgments at study. Within each of these study tasks, 1/3 of the subjects were assigned to each of the three test conditions, recall, generate, and generate + recognize. Thus the design was a 2 (study task) × 3 (test type) between-subject design.

Materials and Procedure

The materials were those used in Experiment 2. The procedure was the same as in the first two experiments up to the point of the test. Test booklets contained separate pages for responses to each category with the category label at the top of the page. For the Recall condition, the labels were provided for the four studied categories with instructions to recall as many of the studied items as possible for each category. The generate and generate + recognize conditions were given four labels corresponding to studied categories and four labels corresponding to non-studied categories. (The counter-balancing of studied and non-studied lists was identical to Experiment 2.) The instructions for the generate condition were to produce the first 15 instances of each category that came to mind. Subjects were told that there were no right or wrong answers, we simply were interested in the first 15 instances that came to mind. The same instructions were given to the generate-recognize along with the proviso to circle any generated items that were recognized as studied items. All tests were self-paced.

Results

Study item performance for each test condition is shown in Figure 3 as a function of study task. The values reflect the mean proportion of the 60 study items produced (or, the case of generate-recognize, produced and recognized) at test. The recall measure (left portion of graph) is the proportion of items correctly recalled. The generate index (center) is the proportion of studied items produced in the generate condition. The right most portion shows the proportion of study items that were both generated and recognized.

Figure 3.

Figure 3

Open in a new tab

The proportion of studied items produced as a function of test type and study task in Experiment 3. Error bars are standard errors.

As can be seen in Figure 3, the effect of study task on recall replicates the results in the first two experiments with pleasantness ratings producing better recall than category judgments. Figure 3 also suggests that the same pattern occurred in the generate + recognize condition. Analysis of the recall and generate + recognize conditions confirmed these impressions. Study task reliably affected memory for list items, F(1, 196) = 105.51, p < .001, ηp2 = .35, but test type had no reliable effect, F < 1, p = .34, nor did it interact with study task, F < 1, p = .50. Thus, the proportion of studied items produced in the recall task and the proportion of studied items produced and recognized in the generate + recognize task was equivalent, with pleasantness ratings leading to higher performance than category judgments in both conditions.

Additional support for the application of a generate-recognize model to these data can be seen by comparing the proportion of studied items recalled, generated, or generated and recognized across the three test conditions. In accord with assumptions of the generate-recognize model, the proportion of studied items produced was affected by test type, F(2, 294) = 92.26, p < .001, ηp2 = .40. Individual comparisons revealed that generate test instructions led to higher output of studied items than both the recall (p < .001) and generate-recognize (p < .001) instructions.

Additional analyses showed that the effect of study task on recall was localized to recognition with no impact on generation. The proportion of generated studied and unstudied target items in the generate and generate + recognize conditions is shown in Table 1. Analysis of these data as a function of study task revealed higher production of studied items than of unstudied (base rate) items, F(1, 196) = 37.32, p < .001, ηp2 = .65, indicating that generation was affected by the study experience. However, neither the study task nor the test type reliably influenced generation, Fs < 1, ps > .46. Thus, the process of generation was not the source of the study task effect on recall. As well, these data show that adding a recognition requirement to the generation task (as in the Generate + Recognize condition) does not appear to have affected the generation process, as indicated by the lack of a test type effect on the proportion of target items generated. Finally, the recognition data shown in the third row of Table 1 were analyzed as a function of study task. Note that these data represent the proportion of generated items that were recognized (and thus the denominator was not 60, but rather was determined for each participant by the number of study items they generated). Consistent with the depiction in Table 1, the analysis showed a reliable effect of study task on recognition, F(1, 98) = 34.79, p < .001, ηp2 = .26, with much better recognition following pleasantness ratings.

TABLE I.

The top row shows the mean proportion of studied and base rate (unstudied) target items produced in the generate test condition. The second row shows the mean proportion of studied and base rate items produced in the generate + recognize test conditions, regardless of whether they were also recognized. The third row is the mean proportion of these generated studied items (in the second row) that were also recognized in the generate + recognize test condition.

Pleasantness Rating Category Judgment
Studied Base Rate Studied Base Rate
Generated on the Generate Test .48 .39 .48 .40
Generated on the Generate/Recognize Test .50 .38 .50 .39
Recognized on the Generate/Recognize Test .77 N/A .55 N/A
Open in a new tab

Note: The standard error was .01 in all cases except for the studied items recognized on the generate/recognize test following pleasantness rating, for which the standard error was .03.

Discussion

The data were consistent with assumptions of a generate-recognize model. The effect of the study task manipulation on performance in the generate + recognize condition was the same as that in the recall condition, consistent with the use of a generate-recognize strategy in recall. Further, production of study items in the generate condition exceeded that of the recall condition, which in turn was higher than the proportion of items produced and recognized in the generate + recognize condition. This is consistent with the assumption that if a generate-recognize strategy occurs in recall, old item output in recall should never be higher than that for generation alone because the addition of a recognition process to generation in recall can only lower performance due to failure to recognize a generated item. In contrast to earlier renditions of generate-recognize theory, Jacoby and Hollingshead’s (1990) version assumes that the generation process is influenced by particular prior experiences, an assumption consistent with the higher generation of studied than unstudied (base rate) items in the generate and generate+ recognize conditions.

Within the generate-recognize strategy, the effect of distinctive processing on recall appears to be primarily on recognition. Study task had no reliable effect on generation of old items. The lack of a study condition effect on generation is almost certainly due to the study conditions used here. Given the test requirement to generate instances to a category label, the use of categorized study lists and semantically focused orienting tasks virtually ensures that the relational information necessary to use the labels to access studied items will be encoded. In other words, the lack of a study task effect on generation almost certainly resulted from equivalent encoding of relational information in the two conditions. On the other hand, recognition of generated items was much higher following pleasantness rating than following category sorting. The addition of item specific information to the shared relational processing yields distinctive processing that allows for precise discrimination between targets and distractors, which has been shown to enhance recognition in previous research (e.g., Hunt, 2003; Hunt & Einstein, 1981).

General Discussion

Because of the assumption that free recall tests do not evoke a signal for familiarity, theoretical descriptions free recall did not include automatic processes akin to the role familiarity plays in recognition. Consequently, McCabe et al.’s (2011) report that automatic processes influence free recall presented an important theoretical challenge to the extant understanding of retrieval in recall. In light of the theoretical shift recommended by McCabe et al.’s results, the research reported here was designed to replicate and extend their findings to an analysis of distinctive processing. Our results fully replicated the data reported by McCabe et al., but additional data from the extension of their design led us to a different theoretical conclusion.

McCabe et al.’s (2011) argument for automatic influences on retrieval is based on their application of Jacoby’s (1991) process dissociation theory to free recall. Process dissociation theory argues that recollection and automatic processes are independent influences on retrieval, and the success of McCabe et al.’s application of the theory implies that automatic processes play a role in output from free recall. Our results nicely replicated McCabe et al.’s findings, showing that divided compared to full attention at study impaired later recollection but had no reliable effect on automatic uses of memory. The effect of the study tasks, however, introduced a potential problem for the application of the theory to category cued recall in that the theoretical estimates were inversely related across the two study tasks. Recollection was higher for pleasantness rating but automatic processing was lower. Jacoby (1998) has identified this pattern as potentially signaling a violation of independence between the processes, which is a fundamental assumption of the theory.

To diagnose the implications of these data, the second experiment added base rate conditions not included in McCabe et al. (2011) or in our first experiment. Jacoby (1998) has clearly described the importance of base rates for ascertaining the validity of his theory in any given situation. Using those guidelines as a diagnostic procedure led to the conclusion that the two-process independence model could not be used to analyze our recall data. Most striking was the finding in Experiment 2 that estimates of automatic processing for the pleasantness rating task were below base rates. This result is nonsensical in the context of Jacoby’s (1991) theory. Experiment 2 also showed that base rates under exclusion instructions were lower than those under inclusion instructions. This result also suggests that assumptions underlying the independence model were violated, but at the same time points toward an alternative in the form of generate-recognize theory (Jacoby, 1998). The third experiment tested the application of Jacoby and Hollingshead’s (1990) generate-recognize model to our paradigm. Unlike the process dissociation theory of independent recollection and automatic processing, the generate-recognize model successfully predicted the effect of the orienting tasks on free recall.

Characterizing Retrieval in Free Recall

Following McCabe et al. (2011), we take “free recall” to entail recall of multiple items to a single cue with no constraint on output order. Although retrieval in free recall is likely an opportunistic process with multiple strategies available to maximize efficiency and accuracy, our data conform nicely to the revised generate-recognize model as do results from preparations similar to ours (Guynn et al, 1999; Guynn et al., 2014). The generate-recognize strategy is well adapted for free recall in that typical requests for free recall not only have multiple correct responses but are also likely to elicit multiple familiar but incorrect responses. For example, “what did you do yesterday?”, “who was at the faculty meeting?”, “recall all of the Birds” are cues associated with many prior experiences. Strategically monitoring the memories brought to mind by such cues can enhance accuracy of output by rejecting incorrect items. The same strategy could be applied to cued recall with only one correct response but in this case may be less likely due to the efficiency of direct access to a single response. For example, the cue “Who was that guy we had coffee with yesterday?” may or may not retrieve a correct memory but is not likely to produce multiple memories. The bulk of the research used to develop and test direct access models, be they dual process (Jacoby, 1991) or single process (e.g., Gillund & Shiffrin, 1984), has relied on cued recall of a single correct response for each cue.

Brainerd and Reyna (2010; Brainerd, Reyna, & Howe, 2009) proposed a hybrid model of recall that captures our point about the difference between retrieval in free recall and cued recall of a single response. The model is a 3-stage Markov process where, following study, items move from an unlearned state to a partial-recall state and with further learning move from partial-recall to a perfect-recall state where the probability of correct recall is 1. The theory proposes that traces in the partial-recall state support non-recollective retrieval whereas traces in the perfect-recall state support recollective retrieval. Recollective retrieval is assumed to be a controlled direct access process, comparable to the concept of recollection in PDP. Non-recollective retrieval essentially is the older version of generate-recognize theory where a cue leads to generation from semantic memory followed by recognition of studied items. Of particular interest here is the application of the theory to free recall versus associative cued recall. Estimates obtained from the model indicate that cued recall of a single item is largely the product of recollection (direct access) whereas free recall is the result of non-recollective (generate-recognize) retrieval (Brainerd & Reyna, 2010). This outcome converges with our conclusion based on very different procedures and models that cued recall of single items is well-described by a direct access retrieval whereas free recall is a generate-recognize process.

The controversial claim made by McCabe et al. (2011) is that retrieval from free recall is influenced by an automatic, familiarity-like process that is independent of recollection. Although our data are not favorable to an independent process direct retrieval model in free recall, they do not eliminate a possible role for familiarity in free recall. This is so because the separate processes of generation and recognition in free recall have been under-analyzed. For example, the generation component has long been taken to be an automatic response to a particular cue (e.g., Bodner, Masson, & Caldwell, 2000; Jacoby & Hollingshead, 1990; Watkins & Gardinier, 1979), but research now has shown that generation operates differently under direct and indirect memory instructions (Mulligan, 2012; Weldon & Colston, 1995) raising the as yet unexplored possibility that generation is subject to controlled as well as automatic influences. Likewise, the recognition decision in a generate-recognize strategy could be affected by familiarity, an explicit assumption of Brainerd and Reyna’s (2010) theory. In fact, the recognition component might be best described by the dual process model with independent contributions of recollection and familiarity. In sum, the generate-recognize processes, while providing a good theoretical account of free recall, may be more complicated psychologically than is now assumed.

Implications for Distinctive Processing

It is well established that encoding differences among items in the context of encoded similarity among the items facilitates recall relative to encoding similarity or difference alone (Hunt, 2012, for a review). Less is known about the locus of this effect in retrieval. Our data partially fill this gap in situations involving a generate-recognize output strategy by showing that the condition identified with distinctive encoding (pleasantness rating of categorized lists) produced better recognition than a control condition whose encoding was focused on similarity (category sorting of categorized lists). Generation of studied items was equivalent for distinctive and organizational processing, a result that is not surprising given that the categories were blocked at study. We assume that the categorical relationships were processed spontaneously with a pleasantness rating study task, as well as in the category judgment condition where the study task was redundant with the spontaneous relational processing. Thus in both conditions the instances were processed in relation to the category at study and were equally accessible in the presence of the category label cue at test. Under conditions modelled by Experiment 3, correct responses are just as likely to come to mind following similarity processing as they are following more detailed distinctive processing, but correct recognition of the accessed items is much more probable following distinctive processing.

Under different circumstances it is quite likely that distinctive processing will facilitate generation. Specifically processing difference in the context of similarity will facilitate generation in comparison to a condition processing only difference. For example if the similarity among the items is not obvious, as is the case with ad hoc categories (Hunt & Einstein, 1981), then a category judgment task would be predicted to produce better performance than a pleasantness rating task. Ad hoc categories encourage encoding of item-specific information, and the category judgment task adds the element of similarity in encoding. Thus in this case the category judgment condition would lead to distinctive encoding, and our prediction is that the advantage in recall from distinctive encoding would be due to higher generation.

Arguably this situation was created by Guynn et al. (2014) who used categorized lists but did so under conditions that likely discouraged processing of the categorical similarity in the item-specific condition. Specifically half of the studied items from each category were atypical, low frequency instances, and item presentation was random with respect to categories in the item-specific encoding task. Both of these factors would reduce the chance for spontaneous detection of categorical relations. Furthermore, half of the items were presented as anagrams in the item-specific condition, again potentially deflecting attention from the categorical structure. In contrast, in the relational encoding condition the study lists consisted only of words blocked by category, and half of the items were sorted into their appropriate category. These conditions assure processing of the categorical relationship among the items. Subsequent recall was better following category sorting than following the anagram task. In other words, in the Guynn et al. study relational processing produced better recall of a categorized list than did item-specific processing, a result that is at odds with all other reports in the literature (Hunt, 2012, for a review). Moreover, the difference between the conditions was traced to higher generation following the relational task. Very likely, the conditions of study in Guynn et al.’s item-specific task discouraged processing of relational information which is necessary for distinctive processing and essential for generation. The point is that whether distinctive processing facilitates generation or recognition will likely depending on the specific comparison conditions, and the degree to which those conditions engender relational processing.

Conclusion

The experiments reported here replicated the important results of McCabe et al. (2011), but additional data from our experiments led to a very different conclusion. In contrast to McCabe et al., our data suggest that free recall in the form of recall of multiple responses to a single cue is not explained by a two independent process model of retrieval. Rather, performance in our studies is best captured by a dependence model of generation followed by recognition. In the context of that model, distinctive processing clearly enhances the recognition component of output in our experiments. However, under different circumstances it is quite likely that generation also would be affected by distinctive processing. On the interesting and important question raised by McCabe et al. of whether there are automatic influences in free recall, our results do not rule out a role for a familiarity-like process in recall. Indeed such a process may be an integral component of recognition in the generate-recognize process, a prospect ripe for future research and theoretical development.

Acknowledgments

The research was supported in part by Grant AGO34965 from the National Institute on Aging to R. E. S. Thanks to the following for assistance in conducting the experiments: Ryan Abel, Brandon Nelson, Alan Hernandez, Josh Brunsman, Julie Niziurski, Nadia Khoja, Katrina Presswood, Marisa Aragon, Joe Tidwell, Ross DeForest, Amy Murray, and Brittany Murray. A subset of the data in Experiment 1 was included in Marisa Aragon’s master’s thesis.

References

  1. Aggleton JP, Brown MW. Thanks for the memories: Extending the hippocampal-diencephalic mnemonic system. The Behavioral and Brain Sciences. 1999;22:471–489. [PubMed] [Google Scholar]
  2. Atkinson RC, Juola JF. Search and decision processes in recognition memory. In: Krantz DH, Atkinson RC, Luce RD, Suppes P, editors. Contemporary developments in mathematical psychology: Vol. 1. Learning, memory & thinking. Freeman; San Francisco: 1974. [Google Scholar]
  3. Bodner BE, Masson MEJ, Caldwell JI. Evidence for a generate-recognize model of episodic influence on word-stem completion. Journal of Experimental Psychology: Learning, Memory, & Cognition. 2000;26:267–293. doi: 10.1037//0278-7393.26.2.267. [DOI] [PubMed] [Google Scholar]
  4. Brainerd CJ, Reyna VA. Recollective and non-recollective recall. Journal of Memory and Language. 2010;63:425–445. doi: 10.1016/j.jml.2010.05.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Brainerd CJ, Reyna VF, Howe ML. Trichotomous processes in early memory development, aging, and neurocognitive impairment: A unified theory. Psychological Review. 2009;116:732–783. doi: 10.1037/a0016963. [DOI] [PubMed] [Google Scholar]
  6. Cermak LS, Verfaillie M, Butler T, Jacoby LL. Fluency versus conscious recollection in the word completion performance of amnesiac patients. Brain & Cognition. 1993;20:367–377. doi: 10.1016/0278-2626(92)90027-j. [DOI] [PubMed] [Google Scholar]
  7. Cohn M, Moscovitch M, Davidson PSR. Double dissociation between familiarity and recollection in Parkinson’s disease as a function of encoding tasks. Neuropsychologia. 2010;48:4142–4147. doi: 10.1016/j.neuropsychologia.2010.10.013. [DOI] [PubMed] [Google Scholar]
  8. Craik FIM. Selective changes in encoding as a function of reduced processing capacity. In: Klix R, Hoffman J, van der Meer E, editors. Coding and knowledge representation: Processes and structures in human memory. Amsterdam; Elsevier, North Holland: 1982. pp. 152–161. [Google Scholar]
  9. Craik FIM, McDowd JM. Age differences in recall and recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1987;13:474–479. [Google Scholar]
  10. Curran T, Hintzman DL. Violations of the independence assumption in process dissociation. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1995;21:531–547. doi: 10.1037//0278-7393.21.3.531. [DOI] [PubMed] [Google Scholar]
  11. Dankert SL, Craik FIM. Does aging affect recall more than recognition memory? Psychology and Aging. 2013;28:902–909. doi: 10.1037/a0033263. [DOI] [PubMed] [Google Scholar]
  12. Dobbins IG, Kroll NEA, Yonelinas AP, Liu Q. Distinctiveness in recognition and free recall: The role of recollection in rejection of the familiar. Journal of Memory and Language. 1998;38:381–400. [Google Scholar]
  13. Dorfman J, Kihlstrom JF, Cork RC, Misiaszek J. Priming and recognition in ECT-induced amnesia. Psychonomic Bulletin & Review. 1995;2:244–248. doi: 10.3758/BF03210964. [DOI] [PubMed] [Google Scholar]
  14. d’Ydewalle G, Van Damme I. Memory and the Korsakoff syndrome: Not remembering what is remembered. Neuropsychologia. 2007;45:905–920. doi: 10.1016/j.neuropsychologia.2006.08.025. [DOI] [PubMed] [Google Scholar]
  15. Eich E. Memory for unattended events: Remembering with and without awareness. Memory & Cognition. 1984;12:105–111. doi: 10.3758/bf03198423. [DOI] [PubMed] [Google Scholar]
  16. Gillund G, Shiffrin RM. A retrieval model for both recognition and recall. Psychological Review. 1984;91:1–67. [PubMed] [Google Scholar]
  17. Graf P, Squire L, Mandler G. The information that amnesic patients do not forget. Journal of Experimental Psychology. Learning, Memory, and Cognition. 1984;10:164–178. doi: 10.1037//0278-7393.10.1.164. [DOI] [PubMed] [Google Scholar]
  18. Guynn MJ, McDaniel MA. Generate-sometimes recognize, sometimes not. Journal of Memory and Language. 1999;41:398–415. [Google Scholar]
  19. Guynn M, McDaniel M, Strosser G, Rameriez J, Castleberry E, Arnett K. Relational and item-specific influences on generate-recognize processes in recall. Memory & Cognition. 2014;42:198–211. doi: 10.3758/s13421-013-0341-6. [DOI] [PubMed] [Google Scholar]
  20. Hirst W, Johnson MK, Kim JK, Phelps EA, Risse G, Volpe BT. Recognition and recall in amnesics. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1986;12:445–451. [PubMed] [Google Scholar]
  21. Hirst W, Johnson MK, Phelps EA, Volpe BT. More on recognition and recall with amnesia. Journal of Experimental Psychology: Learning, Memory, & Cognition. 1988;14:758–762. doi: 10.1037//0278-7393.14.4.758. [DOI] [PubMed] [Google Scholar]
  22. Horton KD, Wilson DE, Vonk J, Kirby SL, Nielsen T. Measuring automatic retrieval: A comparison of implicit memory, process dissociation, and speeded response procedures. Acta Psychologica. 2005;119:235–263. doi: 10.1016/j.actpsy.2005.01.004. [DOI] [PubMed] [Google Scholar]
  23. Humphreys MS, Bain JD, Pike R. Different ways to cue a coherent memory system: A theory for episodic,semantic and procedural tasks. Psychological Review. 1989;96:208–233. [Google Scholar]
  24. Huppert FA, Piercy M. Recognition memory in amnesic patients: Effect of temporal context and familiarity of material. Cortex. 1976;12:3–20. doi: 10.1016/s0010-9452(76)80024-x. [DOI] [PubMed] [Google Scholar]
  25. Huppert FA, Piercy M. Recognition memory in amnesic patients: A defect of acquisition? Neuropsychologia. 1977;15:643–652. doi: 10.1016/0028-3932(77)90069-0. [DOI] [PubMed] [Google Scholar]
  26. Hunt RR. Two contributions of distinctive processing to accurate memory. Journal of Memory and Language. 2003;48:811–825. [Google Scholar]
  27. Hunt RR. The concept of distinctiveness in memory research. In: Hunt RR, Worthen JB, editors. Distinctiveness and memory. Oxford University Press; New York: 2006. pp. 1–25. [Google Scholar]
  28. Hunt RR. Distinctive processing: The co-action of similarity and difference in memory. In: Ross BH, editor. The psychology of learning of motivation. Vol. 56. Elsevier; Oxford: 2012. pp. 1–46. [Google Scholar]
  29. Hunt RR. Precision in memory through distinctive processing. Current Directions in Psychological Science. 2013;22:10–15. [Google Scholar]
  30. Hunt RR, Einstein GO. Relational and item-specific information in memory. Journal of Verbal Learning and Verbal Behavior. 1981;20:497–514. [Google Scholar]
  31. Hunt RR, Rawson KA. Knowledge affords distinctive processing in memory. Journal of Memory and Language. 2011;65:390–405. [Google Scholar]
  32. Hunt RR, Smith RE, Dunlap KD. How does distinctive processing reduce false memory? Journal of Memory and Language. 2011;65:378–389. doi: 10.1016/j.jml.2011.06.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Jacoby LL. A process dissociation framework: Separating automatic from intentional uses of memory. Journal of Memory and Language. 1991;30:513–541. [Google Scholar]
  34. Jacoby LL. Invariance in automatic influences of memory: Toward a user's guide for the process-dissociation procedure. Journal of Experimental Psychology. Learning, Memory, and Cognition. 1998;24:3–26. doi: 10.1037//0278-7393.24.1.3. [DOI] [PubMed] [Google Scholar]
  35. Jacoby LL. Ironic effects of repetition: Measuring age-related differences in memory. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1999;25:3–22. doi: 10.1037//0278-7393.25.1.3. [DOI] [PubMed] [Google Scholar]
  36. Jacoby LL, Begg IM, Toth JP. In defense of functional independence: Violations of assumptions underlying the process-dissociation procedure? Journal of Experimental Psychology. Learning, Memory, and Cognition. 1997;23:484–495. [Google Scholar]
  37. Jacoby LL, Hollingshead A. Toward a generate/ recognize model of performance on direct and indirect tests of memory. Journal of Memory & Language. 1990;29:433–454. [Google Scholar]
  38. Jacoby LL, Toth JP, Yonelinas AP. Separating conscious and unconscious influences of memory: Measuring recollection. Journal of Experimental Psychology: General. 1993;122:139–154. [Google Scholar]
  39. Koriat A, Feuerstein N. The recovery of incidentally acquired information. Acta Psychologica. 1976;40:463–474. [Google Scholar]
  40. Mandler G. Recognizing: The judgment of previous occurrence. Psychological Review. 1980;87:252–271. [Google Scholar]
  41. McCabe DP, Roediger HL, Karpicke JD. Automatic processing influences recall: Converging evidence from the process dissociation procedure and remember-know judgments. Memory & Cognition. 2011;39:389–402. doi: 10.3758/s13421-010-0040-5. [DOI] [PubMed] [Google Scholar]
  42. Mulligan NW. Differentiating between conceptual implicit and explicit memory: A crossed double-dissociation between category-exemplar production and category-cued recall. Psychological Science. 2012;23:404–406. doi: 10.1177/0956797611433335. [DOI] [PubMed] [Google Scholar]
  43. Mulligan NW, Stone M. Attention and conceptual priming: Limits on the effects of divided attention in the category-exemplar production task. Journal of Memory and Language. 1999;41:253–280. [Google Scholar]
  44. Newell BR, Andrews S. Levels of processing effects on implicit and explicit memory tasks: Using question position to investigate the lexical-processing hypothesis. Experimental Psychology. 2004;51:132–144. doi: 10.1027/1618-3169.51.2.132. [DOI] [PubMed] [Google Scholar]
  45. Parkin AJ, Reid TK, Russo R. On the differential nature of implicit and explicit memory. Memory & Cognition. 1990;18:507–514. doi: 10.3758/bf03198483. [DOI] [PubMed] [Google Scholar]
  46. Quamme JR, Yonelinas AP, Kroll NEA, Sauve MJ, Knight RT. Recall and recognition in mild hypoxia: Using covariance structural modeling to test competing theories of explicit memory. Neuropsychologia. 2004;42:672–691. doi: 10.1016/j.neuropsychologia.2003.09.008. [DOI] [PubMed] [Google Scholar]
  47. Richardson-Klavehn A, Gardiner JM, Ramponi C. Level of processing and the process-dissociation procedure: Elusiveness of null effects on estimates of automatic retrieval. Memory. 2002;10:349–364. doi: 10.1080/09658210244000180. [DOI] [PubMed] [Google Scholar]
  48. Toth JP, Reingold EM, Jacoby LL. Toward a redefinition of implicit memory: Process dissociations following elaborative processing and self-generation. Journal of Experimental Psychology: Learning, Memory and Cognition. 1994;20:290–303. doi: 10.1037//0278-7393.20.2.290. [DOI] [PubMed] [Google Scholar]
  49. Tsivilis D, Vann S, Denby C, Roberts N, Mayes AR, Montaldi D, et al. A disproportionate role for the fornix and mammillary bodies in recall versus recognition memory. Nature Neuroscience. 2008;11:834–842. doi: 10.1038/nn.2149. [DOI] [PubMed] [Google Scholar]
  50. Tulving E. Memory and consciousness. Canadian Psychology. 1985;26:1–12. [Google Scholar]
  51. Van Overschelde JP, Rawson K, Dunlosky J. Category norms: An updated and expanded version of the Battig and Montague (1969) norms. Journal of Memory and Language. 2004;50:289–335. [Google Scholar]
  52. Watkins MJ, Gardiner JM. An appreciation of generate-recognize theory of recall. Journal of Verbal Learning and Verbal Behavior. 1979;18:687–704. [Google Scholar]
  53. Weldon MS, Colston HL. Dissociating the generation stage in implicit and explicit memory tests: Incidental production can differ from strategic access. Psychonomic Bulletin & Review. 1995;2:381–386. doi: 10.3758/BF03210976. [DOI] [PubMed] [Google Scholar]
  54. Wixted JT, Squire LR. The role of the human hippocampus in familiarity-based and recollection-based recognition memory. Behavioural Brain Research. 2010;215:197–208. doi: 10.1016/j.bbr.2010.04.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Yonelinas AP. The nature of recollection and automaticity: A review of 30 years of research. Journal of Memory and Language. 2002;46:441–517. [Google Scholar]
  56. Yonelinas AP, Jacoby LL. The process-dissociation approach two decades later: Convergence, boundary conditions, and new directions. Memory & Cognition. 2012;40:663–680. doi: 10.3758/s13421-012-0205-5. [DOI] [PubMed] [Google Scholar]
  57. Yonelinas AP, Kroll NE, Quamme JR, Lazzara MM, Sauve MJ, Widaman KF, et al. Effects of extensive temporal lobe damage or mild hypoxia on recollection and automaticity. Nature Neuroscience. 2002;5:1236–1241. doi: 10.1038/nn961. [DOI] [PubMed] [Google Scholar]

RESOURCES