«The present study investigated whether subjects were sensitive to negative transfer and proactive interference (PI) at encoding and retrieval and ...»
The motivation for the 3 experiments reported here was to investigate how people know to exert control to combat negative transfer and PI. Consistent with the dominant view of monitoring and control in the metacognitive literature (e.g., Nelson & Narens, 1990), the negative transfer/PI sensitivity hypothesis proposed that people monitor for the experience of negative transfer or PI. If they detect negative transfer or PI, they exert control processes to resolve it. Thus, the ability to overcome PI relies on accurate monitoring. The present research focused on three primary questions: 1) Are people sensitive to negative transfer at encoding? 2) Are people sensitive to PI at retrieval? 3) If so, does sensitivity to negative transfer and PI vary with working memory ability? Below I address each question by first summarizing the results and then discussing the theoretical implications for monitoring and control. Because individual differences in WM have been shown to influence PI susceptibility (Kane & Engle, 2000; Rosen & Engle, 1998), WM was included in the analyses to help clarify the results. Thus, I addressed the 3rd question within the two larger sections: sensitivity to negative transfer and PI.
Are people sensitive to negative transfer at encoding?
As mentioned previously, few studies (Diaz & Benjamin, 2005; Eakin, 2005;
Maki, 1999; McGuire & Maki, 2001; Metcalfe et al., 1993; Schreiber, 1998; Schreiber & Nelson, 1998) have investigated whether subjects use the presence of interference as a basis for metacognitive judgments. Only two of these studies (Diaz & Benjamin, 2005;
Metcalfe et al., 1993) have induced PI in their subjects, and only three had subjects make judgments prior to any recall attempts (Diaz & Benjamin, 2005; Schreiber, 1998;
Schreiber & Nelson, 1998). Consequently, very little research has been done on whether people might monitor for negative transfer at encoding. Part of the contribution of the present research was to begin investigating this question.
In short, subjects seemed to be sensitive to negative transfer at the list level but not at the item level, at least when making explicit judgments. Consistent with their recall performance, controls gave higher JOLs than did interference subjects, and interference subjects gave higher JOLs to list 1 than list 2. In E2, this pattern was found even after the immediate recall test of list 1 was eliminated, indicating that interference subjects’ lower JOLs were not due to the experience of retrieval fluency during list 1 recall. Inconsistent with their recall, however, interference subjects gave interference and non-interference items equivalent JOLs and study time, indicating they were unable to locate a more specific source of negative transfer than the entire list. Furthermore, subjects’ list-level sensitivity to negative transfer appeared to be restricted to explicit judgments, as indicated by the equal amount of time spent by controls and interference subjects when they were allowed to pace themselves during study in E3. The results were inconclusive as to whether the presence of negative transfer impaired the relative accuracy of subjects' JOLs: In E1 and E3, subjects were unable to discriminate between items that would later be recalled versus not recalled, but in E2, subjects were able to do so.
The present experiments suggest that subjects were sensitive to negative transfer at the list level but not at the item level. If people are restricted to only a list-level awareness, this could restrict how they initiate control processes. For inhibition, the inability to locate a more specific source of negative transfer suggests that all irrelevant information, regardless of whether it is causing interference, needs to be suppressed.
Although consistent with previous research (e.g, Bjork, 1970; Postman et al., 1968), suppression comes with a later retrieval cost for items that are suppressed (Rosen & Engle, 1998). Consequently, it would be more advantageous to be selective in targeting what is suppressed. For more overt strategies, such as rehearsal, an item-level awareness of negative transfer might be less necessary for instituting appropriate control changes. If overall difficulty in encoding is detected, then this should be sufficient in prompting subjects to switch to a better strategy (see Sahakyan et al., 2004).
Hasher, Zacks, and their colleagues (Hasher et al., 2007; Hasher, Zacks & May,
1999) proposed 3 functions for inhibitory control: access, restraint, and deletion. Of the three, access and deletion are the most relevant here. Access referred to people's ability to keep irrelevant information out of the focus of attention before it intrudes. Deletion referred to people's ability to remove irrelevant information from the focus of attention should it intrude, or when it is no longer needed. If people are limited by having only a list-level awareness of negative transfer, then they can only inhibit identified competitors.
If this is the case, then the most advantageous strategy might be to exert control before negative transfer ever happens (i.e., using the access function of inhibition). For example, in Kane and Engle (2000), subjects completed a recall test after each list. The recall test might have served as an implicit cue that the prior-list information was now irrelevant and should be suppressed. In contrast, in the present E2 and E3, the immediate recall test of list 1 was eliminated (and no instruction to disregard list 1 was given), and the typical recall superiority of high spans over low spans was not found.
As previously mentioned, a list-level awareness of negative transfer is consistent with the ostensible role of strategy change in producing directed forgetting effects.
Sahakyan et al. (2004) found that making a global JOL for list 1 had equivalent recall benefits for list 2 as did a forget cue for list 1. Sahakyan et al. argued that the normal tendency for subjects told to remember list 1 is to not evaluate the efficacy of their list 1 strategy. In the present experiments, subjects only were sensitive to list-level negative transfer if they made explicit judgments like JOLs. When they were allowed to control study time and their judgments were thus implicit, controls and interference subjects did not differ in study time, suggesting they engaged in their normal habit of not evaluating their learning.
However, were subjects really insensitive to item-level negative transfer and do subjects really not evaluate at the list level if they are not explicitly prompted to do so?
E3 subjects spent almost twice as long as E1 and E2 subjects in trying to memorize the word pairs. So, even if subjects were not evaluating the efficacy of their strategy (e.g., staring at the word, rehearsal, making a story), it seems unlikely they would spend so long trying to memorize the word pairs if they did not feel they needed to study them longer. However, the reasoning behind subjects' behavior might not be interference because there were no differences in study time between control and interference subjects. But, subjects in multi-list, non-interference situations might spend less time studying list 2. If this is the case, then the lack of a difference between interference subjects' list 1 and list 2 study times might indicate a sensitivity to list-level negative transfer. Likewise, the lack of a difference between interference subjects' study time for interference and non-interference items might indicate a sensitivity to item-level negative transfer. Future studies should include both subject-controlled study time and JOLs with appropriate control groups (two-list control group with low-interference items to assess the effect of learning-to-learn) to further assess whether subjects are sensitive to negative transfer via implicit judgments.
In addition to the effect of learning-to-learn, people's beliefs about interference is potentially important in investigating whether people monitor for negative transfer and PI. In E2, subjects predicted that other people in the interference condition would recall fewer words than the people in the control group (experimenter provided control group estimate). Thus, people might have a belief that the more items a person learns the harder it is to recall information. The present research could then be explained by an interference-belief explanation rather than a sensitivity to interference explanation.
However, as with Koriat et al. (2004), subjects might only be able to correctly predict recall when they can compare different conditions. What was more intriguing was that subjects predicted that interference items would be better recalled than non-interference items, which bolsters the sensitivity to negative transfer/PI explanation. In essence, despite their beliefs, interference subjects gave lower JOLs than did controls.
Furthermore, this raises the question of how their mistaken belief about interference items might have influenced the allocation of study time in E3. I did not assess subjects' beliefs about interference in E3. Thus if E3 interference subjects noticed that certain word pairs consisted of repeated cues and responses, they might have mistakenly believed they would be easier to remember and studied them less than they would have. Moreover, interference subjects might have given higher JOLs to interference items than they would have if they did not have such a belief. Furthermore, the JOLs for interference items might have been even higher if not for subjects' sensitivity to negative transfer.
Working memory. Did sensitivity to negative transfer vary with WM? In E2, WM did not significantly influence recall performance or JOLs. High spans’ JOLs were more accurate in discriminating pairs that would be recalled versus would not be recalled at the delayed test than at the immediate test. In E3, WM did not influence recall performance, study time, or the relative accuracy of implicit monitoring. Thus, span groups did not differ in their monitoring ability. This could suggest that the difference between span groups might be their ability to control negative transfer and not in their ability to monitor for it. This suggestion might seem odd after considering that in the present research span groups did not differ in PI susceptibility at recall. But, as mentioned previously, high and medium spans might not have been suppressing list 1 as they normally would have if they had completed a list-1 recall test. WM might still influence monitoring ability. Future research could use a forget instruction after list 1 to prompt high spans to use suppression.
Are people sensitive to PI at retrieval?
The results for retrieval were more difficult in interpreting than the results for encoding. First, all 3 experiments had contradictory results as to whether interference subjects were sensitive to PI at the item level, despite the 3 experiments using the same procedure for assessing PI sensitivity. Second, for list-level sensitivity to PI, E2 contradicted E1 and E3 results. I first summarize the results, and then I discuss how the present results coincide with previous research.
E1 and E3 suggest that subjects were sensitive to PI at the list level: Consistent with recall, control subjects gave higher DPOK judgments than did interference subjects.
In contrast, in E2, control and interference subjects gave equivalent DPOK judgments, suggesting they were not sensitive to list-level PI. At the item-level, the picture was more complicated: In E1, interference subjects did not differ in their recall of interference and non-interference items. However, in E1, interference subjects gave slightly higher DPOK judgments to interference items than they did to non-interference items (p =.07), indicating they were being misled by cue familiarity. In contrast, in E3, consistent with their recall, interference subjects gave lower DPOKs to interference items than they did to non-interference items, indicating item-level sensitivity to PI. In E2, interference subjects gave equivalent DPOKs to interference and non-interference items, indicating they were not sensitive to PI at the item level but were not misled by cue familiarity either. Controls gave equivalent judgments to the two types of items in all 3 experiments.