Process dissociation procedure

Introduction

Schifrin and Schneider (1977, in Eysenck and Keane, 2005) distinguished between automatic and controlled memory processes. They claimed automatic process is not limited in capacity or requires attention, whereas the opposite is true for controlled processes.

Graf and Schacter (1987) tagged automatic processing as the process underlying implicit memory while conscious processing reflects explicit memory. They described implicit memory as test performance that is facilitated without conscious recollection of previous episode. On the other hand, explicit memory as defined by Schacter (1987) refers to “conscious recollection of recently presented, as expressed on traditional tests of free recall, cued recall and recognition”.

Explicit memory research are dominated by direct tests like free recall, cued recall, and recognition test are tests whereas implicit memory research is dominated by indirect tests such as word stem completion (WSC), word fragment completion (WFC) where participants are required to complete a word with a degraded cue (e.g. Banana, B_ n_a or Ban_ for word stem). This study adopted Schacter and Graf’s (1986) implicit and explicit memory terms. The term priming (the facilitated use of previous information without conscious recollection – Schacter and Graf, 1986) is also used occasionally to describe the type of facilitation brought about by implicit memory.

The dissociation between implicit and explicit memory have been evidenced by cognitive and neuropsychological research (Schacter and Graf, 1986; Sloman, Hayman, Ohta, Law and Tulving, 1988; Gabrieli, Fleischman, Keane, Reminger and Morrell, 1995; and Jacoby, 1983). Findings reported here have come from studies of amnesia patients (Fleischman, Vaidya, Lange and Gabrieli, 1997), attention (Mulligan and Hartman, 1996), and presentation modality (Graf, Shimamura, and Squire, 1985).

Implicit and explicit memory

Studies investigating neurological disorders such as amnesia, multiple sclerosis and epilepsy have evidenced the dissociation by comparing memory test performance of these patients, to that of healthy, control patients. Amnesia patients have normal perceptual, linguistics and intellectual functioning but have inability to remember recent event and new information (Schacter, 1987). So, they are seriously impaired in standard tests of explicit recall and recognition.

These characteristics were supported by findings of Warrington and Weiskrantz (1968, in Schacter, 1987) who reported that while amnesic patients showed excellent retention of studied words when asked to complete a word stem, they failed to recognize the same words in recognition tasks. Graf, Squire and Mandler (1984) also reported similar findings. Amnesia patients in their study showed impairments when they were told to use word stems as cues for previous studied item, i.e. given explicit instruction but their performance when given implicit instructions (to complete with first word that comes to mind), was comparable to healthy patients’.

Also in support was Gabrieli et al. (1995) and Fleischman et al. (1997) also reported similar findings when they studied MS, a patient who had undergone surgery to remove his right occipital lobe for treatment of epilepsy. Gabrieli at al. (1995) compared MS’s performance to that of other participants on perceptual identification and recognition tests. MS showed no evidence of priming but his recognition memory was intact. However, MS showed priming when tested on category production, a different type of implicit test.

Studies of patients with multiple sclerosis have also informed our knowledge of the dissociation. When matched to control participants, these patients showed impaired performance on tests of immediate recall, despite being comparable to control patients in MS age and verbal intelligent (Grant et al. 1984). Scarrabelotti and Carrol (1998) also compared multiple sclerosis (MS) patients to control patients on a word stem completion and a verbal learning test. They found the same amount of priming in MS and control groups but recall of verbal list was significantly lower in MS group when participants were given explicit instructions. On the hand, recognition was intact, suggesting that these patients may not be exhibiting memory dissociation per se. Two sets of findings supporting this notion are reported below.

Blum, Yonelinas, Luks, Newitt, Oh, Lu, Nelson, Goodkin, and Pelletier (2002) reported evidence suggesting that observed effects may be task and disorder effects rather than dissociation effects. Blum et al. examined patients with different subgroups of MS (Primary Progressive (PPMS), Relapsing Remitting (RRMS), and Secondary Progressive (SPMS) MS subgroups). All groups exhibited free recall deficits but only RRMS and SPMS showed normal pattern of priming.

These neurological evidence are also supported by studies of healthy patients. Of these studies are those investigating the effect of attention.

ATTENTION

Attention is another variable which has been shown to evidence dissociation between implicit and explicit memory. Divided attention paradigm is used to study its effect on both types of memory. In divided attention paradigm, attention is manipulated by asking participants to perform a concurrent task (Wolters and Prinsen, 1997).

Graf and Schacter (1987) investigated the effect of interference on both types of memory using unrelated word pairs. The theoretical assumption of these authors was that unrelated words have no unitized representations in long-term memory in the same way that pairs of related words do (e.g. bread and cake). Thus, any effects observed on implicit and explicit memory will be based on new associations formed during study trials and will not be confounded by the possibility of participants’ strategy such as elaborative encoding. Elaborative encoding is the deep processing of information that involves the use of different strategy such as forming association, grouping information or using mnemonics to make information more memorable (Colman, 2006). Graf and Schacter (1987) found interference impaired performance on explicit recall (as shown by cued recall, pair matching and modified free-recall tests) but have no effect on WSC. Sloman, et al. (1988) also reported impaired performance on recognition task when there was interference but found no such effect on primed fragment completion test. Similar findings was also reported by Mulligan and Hartman (1996) who compared word fragment completion and category exemplar tests to their explicit equivalent, word fragment cued recall and category-cued recall test. Mulligan and Hartman (1996) found divided attention at encoding reduced performance on both explicit equivalents but there was no impaired performance on the implicit tests.

In contrast to these authors that have reported no effect of divided attention on implicit memory, Graf and Schacter (1985) and Rajaram, Srinivas and Travers (2001) reported reduced performance on implicit memory when attention was divided. Wolters and Prinsen (1997) suggested that observing an effect of attention on implicit memory may depend on the difficulty of distraction task used and the type of implicit test. This suggestion was supported by their finding that difficult distraction test interferes with performance on implicit memory as it would be expected for explicit memory. Mulligan and Hartman (1996) also reported divided attention effect on implicit test of category exemplar production, suggesting dissociation within implicit tests itself.

Reports of divided attention effect on implicit memory test or lack of it can be accounted for by different methodological procedure used by these authors. Rajaram et al. (2001) used a Stroop task whereas other studies compared an indirect test (usually WSC but WFC was also used) with a direct test. This notion is supported by Wolter and Prinsen’s (1997). When Wolter and Prinsen (1997) investigated the effect of divided attention comparing cued recall test to word stem completion, they reported a detrimental effect of attention on implicit memory. However, when process dissociation was used, this effect disappeared.

Above findings suggest that explicit memory is generally affected by attention whereas finding an effect of attention on implicit memory is dependent on distraction task and methodology adopted. Other variables (e.g. study processing method and presentation modality) have showed clear cut evidence. Presentation modality is discussed below.

PRESENTATION MODALITY

Presentation modality, the mode of stimulus presentation during study or at test is another factor that dissociates implicit and explicit memory. Jacoby and Dallas (1981) found that changing modality of presentation from study (auditory) to test (visual) strongly affects priming effects on word identification but have little effects on recognition performance. This was supported by Graf, Shimamura and Squire’s (1985) findings when they studied priming effects on WSC and cued recall performance. Graf et al. (1085) tested amnesia and control patients on free recall and word completion test where two of the trials were presented visually and the other two auditorily. On the other hand, test was always visually presented. In the free recall condition, across all participants, changing test modality has no effect, suggesting that changing test modality does affect explicit memory. Nevertheless, changing test modality of the word completion test revealed a different finding. In this condition, performance was higher for words that was studied and tested in the same modality (32.9%) in comparison to words studied and tested under different modality (20.9%). Roediger III and Blaxton (1987) also reported that priming was better for visual presentation and it was also higher,even after a week, when the fragments were presented in the same way as study.

Although evidence described above (neurological, attention and presentation modality) support the distinction between implicit and explicit memory, some similarities between boty types of memory have been reported.

Sloman et al (1988), Graf and Mandler (1984) and Schacter and Graf (1986) all reported evidence suggesting similarity between implicit and explicit memory. Sloman et al. (1988) found evidence of linear forgetting in word fragment completion test which is similar to they type of forgetting usually observed in recognition tests. Similarly, increasing length of study list affects both word fragment and recognition test and Graf and Mandler (1984) reported that both types of memory decayed at a similar rate – word stem completion and recognition test decayed after two hours and. Graf and Mandler also reported that both memory types are affected by practice effect and repetition of study lists significantly increase performance on both tests.

Using two associative elaboration tasks (sentence-generation and word generation) to investigate the effects of elaboration on implicit and explicit memory, Schacter and Graf (1986) showed that both types of memory benefits from elaborative processes.

So far, studies have supported or refuted distinctions between implicit and explicit memory and other have shown similarities between them. These findings are accounted for by three popular theories; processing theory (Jacoby, 1983), activation theory (Mandler, 1980; Graf and Mandler 1984) and multiple memory system theory (Tulving, 1983 not referenced yet).

THEORIES OF DISSOCIATION EFFECT

The processing theory (Jacoby, 1983; Roediger and Blaxton, 1987) relies on the assumption that both implicit and explicit memory have different processing demands with implicit assumed to be governed by data-driven processes and explicit by conceptual processes. Conceptually driven processes involve strategies such as elaboration whereas data-driven are governed by information presented in test materials (Schacter, 1987). This account stressed that test such as WSC relies on the similarities between study information and test information whereas tests like free recall relies on elaborative processes. The theory argues that though both may have conceptual makings, they are primarily driven by different processes (Roediger and Blaxton 1987). This theory is supported by Jacoby (1983) who reported that perceptual identification was better at recall if participants had read word pairs (e.g. hot… cold) than when they generated them from antonyms (hot…), where as the opposite was true for recognition test.

In contrast, activation theory claimed that performance observed on implicit memory is a result of a readily available temporary activation of pre-existing representations or knowledge structures (Graf and Mandler 84, Mandler 80). This activation is assumed to occur on its own without any reference to the relevance between test and study and without elaborative processing; therefore, its contribution cannot be referenced to explicit remembering (Schacter, 1987).

Finally, the multiple memory system put forward by Tulving (1983 in Schacter, 1987) explained the dissociation in terms of underlying differences between episodic and semantic memory. Tulving argued performance on implicit memory test such as word stem completion is similar to performance on semantic memory tests, where participants use pre-existing knowledge. Performance on explicit test is however, comparable to that of episodic memory where they are required to remember recent events.

These theories can account for one aspect of the dissociation or another, but there are also findings they are unable to explain.

The processing account can explain study-test interaction effects such as that reported by Jacoby (1983) and Rajaram and Roediger (1993) but it is unable to explain Schacter and Graf’s (1986) findings that implicit memory for newly formed associations depends on some degree of elaboration. Likewise, the memory account is able to explain memory performance observed in patients of neurological disorder but cannot account for why amnesic patients do no show priming for non-words (Schacter, 1987). The activation theory is unable to account for findings of Sloman et al. (1988) that reported evidence of priming over days and weeks (Schacter, 1987) but can account for evidence of rapid decay reported above by Graf and Mandler (1984) and Sloman et al. (1988)

As well as differences between implicit and explicit memory, there is also proposed difference between encoding and retrieval processes in explicit memory, a topic that is briefly discussed below.

DIVIDED ATTENTION IN ENCODING AND RETRIEVAL

Besides the dissociation between explicit and implicit memory, divided attention paradigm has also been used to investigate the difference between encoding and retrieval processes in memory. Theories of forgetting such as encoding specificity principle (Tulving, 1983) implied that retrieval and encoding processes are the similar. Tulving’s encoding specificity principle claimed studied items are encoded in a specific way with reference to the context in which it was studied (Eysenck and Keane, 2005). Successful retrieval is therefore based on matching of appropriate cue to the memory trace. This also means that performance under a memory test is directly related to the similarity between information presented at study and information available at retrieval (Eysenck and Keane, 2005). Tulving’s theory is supported by a research carried out by Roediger and Tulving (1979) where participants were asked to either recall whole study list or partially recall them by asking them to exclude some information. Roediger and Tulving (1979) reported that instruction to exclude part of the list did not improve recall of the remainder, leading them to conclude that successful retrieval depend on the similarity between study and test information. However, this is not always the case, as shown by various studies discussed below.

Baddeley, Lewis, Eldridge, and Thomson (1984) studied the effect of attention on memory using two tasks to reduce attention. They found reduced attention to strongly affect encoding but have little effect on retrieval. This led them to conclude that retrieval processes are automatic. This was later replicated by Craik, Govoni, Naveh-Benjamin and Anderson (1996) using three experimental tasks: free recall, paired-associate learning and recognition memory task. Craik et al. (1996) reported reduced performance in free recall and paired-associate learning when attention was divided but the effect on dividing attention effect at retrieval was not as much as that observed at encoding. Craik et al. (1996) also that though divided attention have little effect of retrieval; there was a substantial increase in concurrent task reaction time at retrieval than at encoding, where reaction time was only relatively slowed. Reports from Naveh-Benjamin, Craik, Perretta and Tonev (2000) also supported Baddeley et al. (1984) and Craik et al.’s (1996) findings. Craik et al. (2000) reported a large reduction in performance when attention was divided at encoding but found no such effect when attention was divided at retrieval. As well as reporting reduced performance at encoding, Naveh-Benjamin et al. (2000) reported an increase in secondary task cost. They reached the same conclusion as Baddeley et al. (1984); that retrieval processes are protected but at a cost of attention resources. Other studies that reached the same conclusion include Naveh-Benjamin, Craik, Gavrilescu and Anderson, (2000) and Naveh-Benjamin, Kilb, and Fisher, T. (2006).

Despite the compelling evidence discussed above, Fernandes and Moscovitch (2000) argued that the effect of divided attention on retrieval depends on the type of distracting tasks used. They claimed that retrieval and encoding processes involve different systems when attention is divided and finding an effect of divided attention at retrieval depends on the concurrent task used. They claimed when attention is divided at retrieval, the concurrent tasks and memory compete for representational systems whereas when attention is divided at encoding, memory and concurrent task compete for general resources. For this reason, to observe an effect at retrieval, distraction task must be competing for the same system (i.e. observing an effect at retrieval depends on the similarity of both memory and distraction task). On the other hand, any concurrent task which uses the same resources as encoding processes would produce a reduced performance at encoding. They supported their claims with findings that dividing attention at encoding produced large reduction regardless of the distracting tasks (digit monitoring, word monitoring, and word recognition) but an effect of divided attention was only observed when a word-based distraction task was used. However, they also reported increased demand of attention resources at retrieval in comparison to encoding processes.

Critically, one problem with Fernandes and Moscovitch study is that their distraction tasks may have caused additional memory load which could account for their findings. This problem was dealt with by Fernandes and Moscovitch (2002) who investigated whether a word based distraction task that involved no memory load would have the same effect on retrieval as was reported in their earlier research. Despite using a divided attention task that does not load memory, they reported similar findings. This was also evidenced by Fernandes and Moscovitch (2003) who also concluded that retrieval processes is protected and obligatory unless a task that shares the same representational system is used. Though studies above have generally leant towards the idea that retrieval process is protected, findings from other experimental design have yielded conflicting result. Using the psychological refractory period (PRP) design, where stimulus onset asynchrony (SOA) of both tasks are varied to separate them, Carrier and Pashler (1995) found that reaction time for cued recall test was slower at shortest SOA than at longest SOA. Carrier and Pashler concluded this mean memory retrieval was postponed for the completion of concurrent task. The exact opposite should have been found is indeed retrieval processes are protected. Rohrer and pashler (2004) also suggested the possible bearing the nature of concurrent could have on retrieval process. Their finding is reported as part of the study’s discussion.

There is no doubt research reported above have immensely informed our knowledge. Critically, however there is a problem with the methodological procedure used in many studies. Jacoby (1991) argued that these studies reported a task dissociation effect rather than a processing effect. Jacoby (1991) claimed that comparing implicit and explicit tests and taking observed values as an indicator of process dissociation does not provide a pure measure of the dissociation as this type of comparison is easily contaminated by conscious processes which may bias observed automatic processes effect. Schacter (1987) explained that when comparing a direct task to an indirect test as done by many of these research, there is no evidence to suggest that participants were not using strategic processes when they are not suppose to or that automatic processes did not contribute to their direct test performances. To counter this problem of conscious contamination, Jacoby (1991) proposed the process dissociation procedure (PDP).

PROCESS DISSOCIATION PROCEDURE (PDP)

The PDP separates contributions of automatic and controlled uses of memory by using recollected information in opposite ways to compare performance under inclusion and exclusion condition. Estimates of implicit and explicit memory are then derived from performances under these conditions.

Under inclusion condition, for example, in a word completion test, participants are instructed to use previously seen words as cues. If however, they cannot remember previously seen words, they are instructed to complete the stems with any that comes to mind. Under exclusion condition however, the instruction is slightly different; participants still need to use previously seen stems but only as a guide because the instruction here is to complete the stem with any word that comes to mind with the exception of the one they have seen (Hudson and Robertson, 2005). Following these instructions, it is assumed that conscious recollection and automatic processes are used under inclusion condition but the probability of completing word stem under exclusion condition results from the use of automatic processes (Curran and Hintzman, 1995). The procedure uses equations summarized below:

Inclusion = R + (1 – R) A…….(1)

Exclusion = (1 – R) A…… (2)

Thus R (explicit memory) = Inclusion – Exclusion

And

A (implicit memory) = Exclusion/(1 – R)

(Curran and Hintman, 1995)

PDP assumed that the probability of conscious recollection and automatic processes in the inclusion condition is the same as the exclusion condition but automatic processes and conscious recollection are considered to be stochastically independent (Curran and Hintzman, 1995) as given by equations 3 and 4.

This procedure has been successfully investigated by comparing it to studies using single task. For example, Jacoby (1991) compared findings from PDP to findings from single tasks and found that PDP does indeed provide a ‘pure’, task dissociation free estimate of implicit and explicit memory. Similarly, Toth, Reingold and Jacoby (1994) also compared PDP findings and those from single tests studies. They reported that task dissociation procedures are contaminated by conscious processes and that PDP showed no self-generation effect or differential effects of semantic or non-semantic processing. Wolters and Prinsen (1997) reported reduced performance on both implicit and explicit memory under different conditions using task dissociation procedure but when they ruled out conscious contamination using process dissociation procedure, they found divided attention affect on explicit but not implicit memory.

Although the procedure seem well supported, some studies (e.g. Curran and Hintzman, 1995; Stern, McNaught-Davis and Barker, 2003; and Horton, Wilson, Vonk, Kirby and Nielsen, 2005) has reported evidence suggesting flaws in the assumptions made by the procedure.

Curran and Hintzman (1995) showed flaws in the procedure’s assumption of independence between inclusion and exclusion under various manipulations. Curran and Hintzman reported direct evidence of correlations between R and A that goes against assumptions of equations 1 – 4. Their result also suggested that participants also assumed strategies other those instructed by the procedure. Horton et al. (2005) also reported that in comparison to speeded response procedure, estimates of automatic memory retrieval from process dissociation procedure were consistently lower with the exception of when experimental condition yielded low conscious retrieval.

Most studies of implicit memory tested the effect of divided attention at encoding. One aim of the present study is to investigate the effect of divided attention on implicit memory when attention is divided at retrieval using a procedure that eliminates conscious contamination. Another aim of this study is to investigate whether retrieval processes in explicit memory is still obligatory even when there is little chance of conscious contamination that may have confounded previous findings and if this asymmetry reported by previous studied exists in implicit memory. These aims are investigated using Jacoby’s (1991) process dissociation procedure. The final aim of this study is to compare performances on implicit and explicit memory.

Using PDP, the study predicts findings similar to that of Mulligan and Hartman (1996). Therefore, this study predicts that dividing attention either encoding or retrieval will have no effect of implicit memory. However, divided attention at encoding is expected to cause detrimental effect of explicit memory whereas divided attention at retrieval is expected to be protected as reported by Craik et al. (1996). Subsequently, time spent during divided attention at retrieval is expected to be significantly higher than that of divided attention at encoding (Craik et al. 2000).

Curran and Hintzman(1995) claimed that assumption of independence claimed by the procedure is violated by participants obtaining a perfect recollection score ( R = 1). They argued that when this happens, estimates of automatic processes is undefined. Equation 4 also becomes problematic if exclusion scores = 0. Curran and Hintzman claimed that in this case, equation 4 becomes invalid because it is no longer scaled by level of R. Jacoby et al (1993) however, proposed that this problem can be overcome by excluding participants with zero exclusion scores. To this end, this study aim to compare estimates of automatic and conscious recollection using original PDP equations as proposed by Jacoby (1991) with adjusted estimates (Jacoby et al., 1993).

Experiment overview

The experiment used the process dissociation procedure (PDP) to examine effect of divided attention (DA) on implicit and explicit memory using word stem completion test (WSC). PDP as explained above, has within it two phases (study and test) and two conditions (inclusion and exclusion). The study phase involved the presentation of 2 randomized lists of 16 words. In the test phase, participants were presented with word stems of the studied lists and another unstudied list, making a total of 48 words.

Within the procedure, the studied lists are assigned to inclusion and exclusion condition whereas half of the unstudied list is assigned to the inclusion condition and the other half to the exclusion condition. The inclusion condition of the procedure requires participant to fill in the word stems with words they have previously seen or with another that comes to mind. On the other hand, the exclusion condition requires them to complete the stems with words that are completely different from what they have studied.

In the experiment, words that were to be included appeared green; words from the excluded list appeared red; and half of C appeared green while the other half appeared red.

There were 3 experimental conditions: full attention (FA), divided attention at encoding (DA-E) and divided attention at retrieval (DA-R).

Method

Participants

60 undergraduate students of University of Lincoln participated either voluntarily or for 1 course credit. Age ranged from 18 to 40; mean age = 20.7.

Design and Materials

The study used a 2 X 2 X 3 design in which Memory processes (implicit and explicit) was the within-subject factor and Attention (FA vs. DA-E vs. DA-R) was manipulated between- subjects. The experiment was fully counterbalanced by having three possible sets of words and the inclusion and exclusion conditions were based on the study set. 20 participants were assigned to each of the attention condition. To maintain counterbalancing, participants were assigned to study sets one after the other. So participants one did study set 1 and 2 did set 2… This order was also followed when assigning participants to experimental condition.

The experimental programme was written to accommodate 3 lists of 16 words which were between 5 and 8 letters long. These lists were classed as ‘inclusion’, ‘exclusion’, and ‘baseline’. Baseline words were words that participants were not exposed to but nevertheless asked to complete their stems (see above for inclusion and exclusion definition). Baseline words were included to further support estimates of automatic and conscious recollection derived from inclusion and exclusion scores.

Words were presented visually, in lower case, at the centre of the screen for approximately 4 seconds. The study phase was therefore, approximately 2 minutes. The first three letters of the words were used as stems. The colour of words in the study phase was black. Above each stem in the test phase was the instruction “FILL IN THE STEM WITH A WORD YOU HAVE JUST SEEN” (for green stems) and ‘FILL IN THE STEM WITH A WORD YOU HAVEN’T SEEN’ (for red stems).All stems have three or more solutions and each stem was unique. So for example, the word SHAPE had the stem SHA and there was no other stem that began with SHA.

The divided attention task was a digit monitoring task. There were two digit monitoring recordings: one of a female’s voice and the other was a male’s. The digits were generated using an online random number generator. Participants’ task was to monitor these recording and detect when a single digit occurred consecutively (e.g. 67891333). Each recording is 5 minutes long with possible 10 detections at encoding for the male voice and 9 for the female voice. For retrieval, if tape was switched at 3.10 minutes, there are 13 detections for male’s voice and 14 for female but if switched at 3.20 minutes, there is an additional detection. The number of possible detection depended on the length of the recording played which in turn depended on the condition and participants. Naturally, participants in the DA-E condition used the same amount of time (roughly 2 minutes).

Procedure

Participants were tested individually and they were all required to read the consent form and sign their names on the consent sheet shows an example) before study began. The basic instruction was for participants to remember the words they are about to be shown. There was a brief pause between study and test for further instructions. For participants in divided attention (DA) conditions, instruction included the performance of the divided attention task.

In DA-E condition, participants performed the digit monitoring task during the study phase,