The false memory phenomenon has been studied in a variety of ways. However, studies pertaining to feature and conjunction errors have become an area of interest for many researchers. For example, Jones and Jacoby give the following example. Suppose a participant studied "buckwheat, blackmail, and jailbird." A common response a participant would give on a recognition test following the study list would be "buckshot" or "blackbird" (Jones & Jacoby, 2001). Thus, the participant makes the mistake of combining parts of old words to form new words. Confusion is quite easy to provoke because "new" words are actually composed of features of previously studied "old" words (e.g. "blackbird" – both words are partly old). Also, it is important to note that a word can be partly old (e.g. "buckshot").
Hence, there are two types of false recognition errors that are made in this kind of study:
DUAL PROCESS THEORY (DPT)!!! What is DPT? Basically, it can be broken down like this... One’s memory has 2 options when engaging in the act of recognizing something. We can 1) remember things in an "automatic" manner, AKA "familiarity" or 2) remember things in a controlled, conscious process, AKA "recollection."
Here is an easy way to understand the differences between "familiarity" and "recollection" processes. Suppose that you saw a movie last night. There was a particular scene that was very funny. You consciously remember the funny actor’s comments. When you think back of this conscious experience, you are engaging in recollection. Familiarity processes, however, are different. For example, you know what your name is. Yet, you do not have a conscious experience of the first time you were called by your name. It is familiar to you – it’s an aspect of life that is an automatic process. We do not consciously think about our names – we just know what we are referred to by others. Thus, this is an example of familiarity. In terms of "cognitive speed," familiarity processes are FAST and automatic and recollection processes are SLOW and strategic.
** THE ROLE OF FAMILIARITY AND RECOLLECTION ON RECOGNITION TESTS**
BOTH familiarity and recognition exert an influence upon hit rates for old words. For instance, if participants studied: JAILBIRD and BLACKMAIL and later saw: JAILBIRD [on the recognition test], one could choose the correct "old" response JAILBIRD by using familiarity and/or recognition. That is, you might consciously recollect the word JAILBIRD and choose it for that reason or LIGHT might merely seem familiar to you and you'd choose it for that reason. For false alarms the situation is different. If you are presented with the word BLACKBIRD on the recognition test recollection guard against falsely recognizing it (i.e. "It wasn't BLACKBIRD it was BLACKMAIL") whereas familiarity will tend to elevate false alarms (i.e. "There's something familiar about BLACKBIRD") (Jones & Jacoby, 2001; Yonelinas, 1994, 1997; Yonelinas, Kroll, Dobbins, Lazzara, & Kingiht, 1998).
There are two predictions mentioned in the present paper with regard to feature and conjunction errors made: 1) feature and conjunction lures are poor retrieval cues for study primes, which hinders recollection and 2) feature and conjunction errors are based on familiarity judgments, so FT and CJT errors posses a greater familiarity than new words (Jones & Jacoby, 2001). So , errors should be common because familiarity will be relatively unopposed by recollection.
The differing theories bring about an important question: What experimental manipulations affect feature and conjunction errors? It has been shown that hit rates can be affected by several manipulations BUT not FT and CJT effects. The few variables that exist for FT and CJT effects are demonstrated in the following: 1) individuals with hippocampal damage – they exhibited higher CJT error rates when a delay between study word primes was minimal (e.g. 1-vs. 5 word lag; Kroll et al., 1996) 2) normal participants produced high levels of CJT errors when shown face stimuli simultaneously compared to face stimuli that were presented sequentially (Reintz & Hannigan, in press).
The present authors use the familiarity-recollection account as a basis for their FT and CJT experiments. Here’s a look at how the experiments are broken down:
EXPT 1 and EXPT 2: a recollection manipulation – familiarity is not manipulated
EXPT 3 and EXPT 4: focus on retrieval manip’s: influence both recollection and familiarity; show that when study items are more memorable, recollection is used to avoid errors. This demonstrates that both familiarity and recollection processes work in opposition when FT and CJT lures are shown during recognition.
** Expectations ** The experiments should influence old-new discrimination but not FT and CJT effects. Little change is expected for FT and CJT effects because of the minimal amount of recollection that should be used.
Experiment 1 draws upon previous research pertaining to divided attention studies. There have been several studies of this nature demonstrating that dividing attention affects recollection, not familiarity (Jacoby, 1996, 1999; Jacoby, Toth & Yonelinas, 1993; Wolters & Prinsen, 1997; see Mulligan & Hartman, 1996 and Schmitter-Edgecombe, 1999, for a further distinction). However, few studies (one to be exact) have applied these findings to FT and CJT errors (Reinitz et al., 1994). Effects were found for old conditions but not for FT and CJT conditions.
PURPOSE: To show that an encoding and retrieval manipulation could produce similar patterns of data by dividing attention during encoding or retrieval. If this manipulation affects old-new discrimination but not FT and CJT effects, then the problem exists in ACCESSABILITY not retrieval and encoding
Method
Participants. 96 New York University undergraduates participated for course credit for intro. to psychology course. One person’s data was discarded because of failure to follow instructions.
Materials and equipment. 258 compound words were used. 240 words were used to form 80 critical sets of triplets. The remaining 18 words were used as a buffer study items or practice test items. A target word (e.g. checkpoint) was contained in each set of triplets (6-11 letters long), and 2 word primes that contained "old" words found in the target word (e.g checklist and needlepoint). 80 triplets were assigned to 1 of 4 lists with 20 triplets each – the lists were matched for mean word length of the targets. The experiment was run on a computer and a stereo was used to play the tape containing a female reading single-digit numbers.
Design and procedure. A 3 (Group: full attention, divided attention at study, divided attention at test) * 4 (Item type: old, feature, conjunction, or new word) mixed design. Attention was a between subjects variable and item type was a within subjects variable. Participants were randomly assigned to 3 groups (containing 32 partc/grp). All groups heard single digit numbers (@ 1.5 s rate). The 3 groups did different tasks for the attention manipulation: #1: tracked single-digits @ study #2 tracked digits @ test #3 did not attend to the digits.
Participants studied 3 lists and were tested over 4 lists. The 4 lists contained different item types: old, conjunction, feature, or new word – and were rotated equally often (See Table 1). Study orders and item types were counterbalanced to ensure equal distribution at test and at study.
Study phase: 1) Participants read words silently in preparation for a memory test. 2) The divided attention at study group: did digit tracking task AND read/studied words -- it was emphasized to do well on digit tracking task.
Digit task: 1) told to say "hit" after every sequence of 3 odd numbers – experimenter said "miss" if sequence was missed.
Divided attention at test and full attention groups: concentrated on reading/studying words, not on digit task. Experimenters ensured that instructions were clear.
Test instructions: given after study phase. Participants were given a practice "yes-no" recognition test before taking the actual test. They were informed of old and new words, but not of FT and CJT lures. Participants were told to respond YES to OLD words and NO to NEW words. They had a 3s time limit to respond.
Divided att’n group: took rec. test and did digit tracking at same time
Divided att’n @ study group: took rec. test and did digit tracking at same time; digit tracking emphasized.
Full att’n: told to ignore digit task when taking rec. test.
Results
For each of the 3 groups, the mean rate of responding "old" to the various items had the following pattern: old words > CJT lures >FT lures >new words (See Fig. 1). Each old response significantly differed. The false alarm rate for the divided attention group was significantly greater than those in the full attention group. This is similar to previous findings (Reinitz et al., 1994). ** d’ scores were also calculated (See extra handout). ** Performance of the 3 groups was important. Dividing attention should affect recollection (ability to discriminate old from new items should decrease). There should also be no change in discrim’n of FT and CJT words from new words. This hypothesis was supported. There was a significant effect for GROUP and ITEM TYPE and a significant interaction. Further analysis revealed the following:
In regards to accessibility, the FULL ATTN group have greater access to study words than DIVIDING ATTN groups did. Hence, recollection should be hindered (more) for those who do not have access to the conscious recollections needed to make a correct rejection (and consequently FT and CJT errors are made).
Experiment 2A and 2B focuses on response deadline manipulations. When response time is decreased, old-new discrim. to test probes is decreased; but not FT and CJT effects. This gives further evidence that recollection is a SLOW process.
PURPOSE: To show that response deadlines have little or no effect on FT and CJT errors.
** This procedure has been modified from using within-subject variable to between subject variable for response deadlines (Jones et al., 2001). **
Method
Participants. Two groups of 16 NYU undergraduates participated (n=32) for credit toward an intro. psyc. course.
Design and Procedure. The experiments used a 2-variable, mixed design: Item type (old, FT, CJT, new word) – manipulated within subjects and Response deadline (short or long) was manipulated between subjects. The materials, lists, and instructions (expect for divided attn) were the same as mentioned in Exp. 1.
Short deadline group (2A) told to respond to word before 850-ms deadline.
Long deadline group (2B) told to wait for response signal to appear below the test word and then respond within 750-ms after the response signal appeared. There was a 1400 ms delay – no responses were allowed during this time. The deadline for group 2B was 2150 ms. The "waiting period" was used in order to encourage more recollection.
Results
As shown in Table 2, hit rates for the long deadline group were higher than those for the short deadline group. Conjunction rates for the long and short groups were equal. Feature error rates for the long and short groups were almost equivalent.
The false alarm rate for new words did not differ between the 2 groups. Corrected FT and CJT rates did not differ between the 2 groups. The results replicate the findings by Jones and colleagues (2001). They provide yet another example that recollection processes are SLOW.
Experiment 3 Experiments 1 and 2 did not provide direct evidence for recollection. Experiment 3 was designed to provide direct recollection evidence.
PURPOSE: To show that 2 processes account for CJT errors. Two manips were used: 1) number of study presentations and 2) response deadlines (short and long).
Repetition should increase both familiarity and recollection (Jacoby, 1999; Jacoby, et. al., 1998; McElree et al., 1999). It is also possible that repetition can increase CJT errors. There are 3 phases in Expt. 3: 1) list of study words presented visually – some presented once, others repeated 2) study words presented aurally, words presented once 3) words that were read or heard were presented in a test list that contained new words. Participants were to exclude visually presented words from old responses. Prior evidence has demonstrated the differences between long deadline and short deadline groups in relation to rates of exclusion.
Long deadline group: showed a decrease in exclusion errors for REPEATED words (v. non-repeated words).
Short deadline group: showed an increase in exclusion errors for repeated words v. non-repeated words.
DIFFERENCE: The current experiment has 1 study list and the test does not involve list discrimination.
Participants were informed of CJT and FT test lures in order to further enhance recollection – unlike Exp. 1 and 2.
Method
Participants. 64 NYU undergraduates participated for course credit in intro. psyc class.
Design and procedure. A 2 (Group: deadline, wait) * 7 (Item type: old word-1 presentation, old word-4 presn, FT condition-1 presn, FT-4, CJT-1, CJT-4, new) mixed design. There were 8 lists of 10 triplets (taken from 80 trips in Exp. 1-2). Six lists corresponded to study items and the remaining 2 lists were used as new items on recognition test. There were 200 critical trials. Item type distribution throughout study list was equivalent. Table 3 shows the spacing of repeated CJT study primes. Each study word was presented 1 or 4 times. Participants read words aloud upon presentation. Deadline procedures were similar to that of Exp. 2A and 2B.
Results
As shown in Table 4, participants recognized old words more after 4 presentations than 1P under the long deadline condition. The same pattern was evident for the short deadline condition. Conjunction error rates were highest after 4P in short deadline condition (but less than old). Feature error rates followed a similar pattern, but were less than CJT rates and OLD rates. There were main effects for Item type and Repetition and the interaction of those factors was significant. Performance for old words was significantly higher for the long deadline group than for the short deadline group. "Control" for CJT and FT errors was not evidenced to be very strong.
Experiment 4 attempted to remedy the weak effects found in Exp. 3 for using recollection to avoid CJT and FT errors.
PURPOSE: To attempt to show (once again) that recollection can be used to avoid making CJT and FT errors by changing the procedure (of Exp. 3) slightly.
Changes: 1) decrease the length of study list – increase recollection by using shorter lists. 2) words in repetition condition were presented 3 v. 4 times. 3) 5 items types: old word-1 presentation, old word-3P, CJT lures with components presented once, CJT-comp-3P, new word lures (no FT lures).
** It was important to increase wait period so recollection could occur (or at least be more likely to occur).
Method
Participants. 40 undergraduates from Victoria University of Wellington participated in a room that contained individual computer stations. They received $5 for their participation.
Materials. As in Exp 1-3, 80 triplets were used. Some words were modified.
Design and procedure. A 2 (Response deadline: short/long) * 5 (Item type: old word-1 presentation, old word-3P, CJT-1P, CJT-3P, new word) design. Response deadlines was manipulated between subjects, but item type was manipulated within subjects.
The 80 triplets were divided into lists of 16 triplets. There were 4 different study and test orders for each study-test block. Participants read each word (at study) to themselves prior to taking the memory test. Example of conjunction lures were given. There were 24 trials on each test. At test, a focal point was presented (+++++) for 500 ms. Then, there was a blank screen for 250 ms before a test word was shown.
Results
Despite changes, the same patterns found in Exp. 3 were found in Exp. 4 (See Table 5). The short deadline group showed a repetition effect for the CJT condition, but the long deadline group did not. False alarm rates differed for the 2 groups. Furthermore, it was found that the increased response deadline DECREASED the probability of correctly recognizing old words. However, study repetition increased the chances of correctly recognizing old words. The repetition effect was significant for the short deadline group, but not for the long deadline group.
Discussion
In summary, the various manipulations used throughout Experiments 1-4 demonstrated the differing ways in which recollection for old, conjoint, feature, and new words is affected -- or in many cases is not affected. Dual process theory accounts for the data but retrieval processes need to be taken into consideration as well (Jones & Jacoby, 2001). Because feature and conjunction items contain high amounts of familiarity, it makes sense that a the more "automatic" familiar memorial process is used when making judgments about conjunction and/or feature stimuli. Yet, in order to correctly REJECT a conjunction and/or feature stimuli, one must be able to accurately recollect which part of the CJT and FT stimuli is old and which is new. But, CJT and FT errors are based on FAMILIARITY – recollection need not be present. Contrastingly, single process theory asserts that familiarity only is used when making decisions about stimuli (e.g. Rubin et al., 1999). However, single process models cannot account for particular dissociations (See p. 96, Jones & Jacoby, 2001). Familiarity alone cannot explain the results found in Experiments 3-4 (which show recollection).
Contrary to previous studies (e.g. Reinitz & Hannigan, in press; Rienitz et al., 1992), the authors from the current paper have found that recollection is NOT used to produce conjunction errors. Recollection is used to REDUCE conjunction errors. There are many arguments pertaining to reasons why conjunction and feature errors are made. Are retrieval cues too inadequate (in relation to the compound words)? Are item information and associative information stored in distinctly different conceptual units? The answers to these questions have not yet been completely addressed.
|
|
|
|
|
|
|
|