The motoric fluency effect on metamemory

Journal of Memory and Language 95 (2017) 116–123

Contents lists available at ScienceDirect

Journal of Memory and Language journal homepage: www.elsevier.com/locate/jml

The motoric fluency effect on metamemory Jonathan A. Susser a, Jennifer Panitz b, Zachary Buchin a, Neil W. Mulligan a,⇑ a b

Department of Psychology and Neuroscience, University of North Carolina at Chapel Hill, 235 E. Cameron Avenue, Chapel Hill, NC 27599-3270, United States Department of Psychology, School of Social Sciences, University of Mannheim, Schloss, Ehrenhof-Ost, 68131 Mannheim, Germany

a r t i c l e

i n f o

Article history: Received 24 May 2016 revision received 6 December 2016

Keywords: Metamemory Motoric fluency Judgments of learning

a b s t r a c t Predictions of future memory are often influenced by the ease or fluency of processing information. Susser and Mulligan (2015) recently demonstrated that motoric fluency (of writing with the dominant or non-dominant hand) may likewise affect these predictions. In the present study, we report five experiments that specify the locus of this motoric fluency effect. In Experiment 1, we examined whether the effect was driven by differences in effective study time across hand conditions. In Experiment 2, we assessed whether the effect could be obtained without any visual feedback from handwriting. In Experiments 3a and 3b, we investigated the contribution of visual feedback alone. In Experiment 4, we used prestudy JOLs to determine whether participants may develop a belief about handedness in the context of the experiment. Taken together, the results indicate that the motoric act of producing information in a fluent or disfluent manner is sufficient to produce an effect on memory predictions, that visual information from writing does not contribute, and that on-line interaction with the task plays a role. The experience of motoric fluency appears to be another cue that affects metamemory. Ó 2017 Elsevier Inc. All rights reserved.

Introduction Assessing how people monitor and predict their learning is a core component of metamemory, and identifying the cues that people use to make their predictions is a primary goal of this research. Recent research on metamemory has focused heavily on the extent to which people incorporate processing fluency (Alter & Oppenheimer, 2009) into their memory predictions. In particular, the ease of perceiving (e.g., Besken & Mulligan, 2014), retrieving (e.g., Benjamin, Bjork, & Schwartz, 1998), and encoding (e.g., Koriat & Ma’ayan, 2005) information have all been associated with greater memory confidence, even in cases where memory performance is not similarly affected (e.g., Benjamin et al., 1998; Besken & Mulligan, 2013, 2014). Much of the research on fluency-based effects in metamemory has focused on perceptual or conceptual fluency upon initially encountering information. For example, Besken and Mulligan (2014) manipulated the perceptual fluency of auditory information by having participants hear either intact or degraded words over headphones. Participants heard a study word, said it out loud, and made a judgment of learning (JOL; a 0–100 confidence rating) regarding its likelihood of recall on a subsequent memory test. Participants were quicker to name the intact words and were also ⇑ Corresponding author. E-mail address: [email protected] (N.W. Mulligan). http://dx.doi.org/10.1016/j.jml.2017.03.002 0749-596X/Ó 2017 Elsevier Inc. All rights reserved.

more confident in their ability to remember them, consistent with the notion that perceptual fluency affects metamemory. Actual recall performance, however, was greater for the more difficultto-perceive degraded items. Other research has similarly assessed the role of fluency as study stimuli are first identified or processed (e.g., Benjamin et al., 1998; Besken & Mulligan, 2013; Undorf & Erdfelder, 2015). After experiencing information, though, we frequently act on it. Research on fluency and metamemory largely focuses on variation in the internal processing of information, or what we might call cognitive fluency, but what of the fluency of motor response, or motoric fluency? Little research has examined the relation of physical or motor actions to metamemory, and still less the effect of the fluency of such actions. Alban and Kelley (2013), motivated by theories of embodied cognition, showed that variation in the physical weight of a held study object can influence JOLs. Although this study speaks to the role of physical actions in metamemory, it does not tap into any effect of fluency. Susser and Mulligan (2015) more directly investigated motoric fluency by having participants read and copy down words on note cards. Participants wrote some of the words with their dominant hand and others with their non-dominant hand, and provided a JOL after each one. As expected, writing times were faster for words written with participants’ dominant hand, validating the manipulation of motoric fluency. More critically, participants’ JOLs were greater for words written with the dominant than

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

non-dominant hand; however, actual memory performance was unaffected. An additional experiment using aggregate, list-end JOLs found the same pattern, while a separate questionnaire-based study revealed that participants did not have a priori beliefs about an effect of writing hand on memory. These results imply that the ease of motor processing affects metamemory, just as the ease of perceptual and cognitive processing does, but questions remain about how exactly this manipulation of motoric fluency produces its effects. Recent research on perceptual and conceptual fluency has further explored the extent to which fluency effects in metamemory are actually due to their purported bases. For example, there is active debate about whether perceptual fluency effects are driven by the manipulation of perceptual processes (e.g., Mueller, Dunlosky, Tauber, & Rhodes, 2014; Susser, Jin, & Mulligan, 2016; cf. Rhodes & Castel, 2008). Indeed, a prerequisite for claiming a fluency-based locus necessitates that a manipulation has a measureable effect on perceptual processing. The font size of study words affects JOLs and was initially thought to do so through perceptual fluency (Rhodes & Castel, 2008). Subsequent research, however, indicated that font size did not actually influence processing measures (Mueller et al., 2014). In contrast, the manipulation of perceptual fluency used by Besken and Mulligan (2014) actually does affect the speed of perceptual identification. Still, even demonstrating that a manipulation affects processing speed in the appropriate modality is insufficient to isolate the cause of the effect. A manipulation may have effects on multiple dimensions of processing and not just on the dimension of interest. As discussed below, in the case of writing hand, one needs to isolate the motor component from other correlated effects of the manipulation. In a similar vein, research on metamemory often distinguishes between contributions of experience and beliefs to JOLs (e.g., Koriat, Bjork, Sheffer, & Bar, 2004). Experienced-based contributions reflect the subjective experience of interacting with materials, whereas beliefs refer to our theories about how memory operates (this distinction is similar to the non-analytic—analytic distinction, Matvey, Dunlosky, & Guttentag, 2001). Hypotheses about fluency and metamemory generally propose that the experience of fluent processing informs metamemory judgments directly (an experience-based or non-analytic contribution to JOLs). In contrast, it is also possible that the effect of a fluency manipulation might be based on a belief about memory (an analytic process), in which the individual expresses the belief that a certain class of items is, on average, more memorable than another class. Finally, it should be noted that belief-based contributions encompass a priori beliefs, in which the individual responds to metamemorial judgments with a preexisting belief about how memory operates, as well as beliefs that develop in the context of a set of memory judgments (e.g., in the context of an experiment on metamemory) (e.g., Mueller, Dunlosky, & Tauber, 2016). The latter case may seem to muddy the distinction between experiential and belief bases; these beliefs develop (or perhaps are triggered) through new experiences (e.g., with materials about which previously the individual had no beliefs). However, these on-line beliefs can be considered beliefs because, once developed, they are applied in a way that differentiates sets of items. The issues relevant to other fluency effects also arise with respect to motoric fluency. It is certainly plausible that the effect of writing hand on JOLs represents an effect of motoric fluency. Supporting this idea, writing hand has a substantial effect on writing times. However, as with the case of manipulations of perceptual fluency, there are other plausible accounts. In the present study, the first several experiments assessed whether the motoric fluency effect is actually due to the motoric component of the manipulation or due to other factors implicated by this manipula-

117

tion. The last experiment investigated the motoric fluency effect with respect to beliefs versus experience. The results of Susser and Mulligan (2015) are certainly consistent with the idea that the motor component of the manipulation drives the effect in metamemory, but several alternative possibilities exist. The first has to do with the details of how the motoric fluency manipulation was implemented. Specifically, in Susser and Mulligan’s experiments, the study-trial duration for each word was 13 s, which included the writing time (but not the JOL judgement). Therefore, participants had more time to study the word after writing in the dominant condition because these words were written much more quickly (in an average of 4.64 s compared to 7.82 s in the non-dominant condition). If participants were sensitive to the discrepancy in effective study time, they may have used this information in making their JOLs. In essence, the motoric fluency effect might be due to participants’ beliefs about the effect of study time (see Koriat & Ma’ayan, 2005) rather than the effect of motoric fluency itself. To assess this possibility, Experiment 1 equated effective study time for all items by keeping postwriting times constant. Another important issue arises from the research of Briñol and Petty (2003). These researchers documented an effect of handwriting (with the dominant or non-dominant hand) on affective judgements and suggested that the result might actually be due to the visual appearance of the writing as opposed to the feeling of motoric fluency or disfluency. In particular, writing with the nondominant hand produces text that looks shaky and unclear, which may reduce confidence in that information. This idea represents a form of perceptual feedback that might account for the effect. Based on this proposal, it is important to examine the potential role of perceptual (visual) feedback in the (purportedly) motoric fluency effect. We approached this goal from two directions. In Experiment 2, we examined whether the motoric fluency effect is obtained in the absence of visual feedback by having participants copy down words without being able to see any visual product. In other words, is the motoric component alone enough to inform JOLs? In Experiments 3a and 3b, we did the inverse: we examined whether presenting participants with the visual product of fluent or disfluent writing (by using the handwriting of other subjects) – without participants writing anything themselves – could produce the effect. Finally, Experiment 4 assessed whether on-line beliefs contribute to this fluency effect. The results of the questionnaire study in Susser and Mulligan (2015) imply that the motoric fluency effect is not due to an a priori belief, but this does not rule out the possibility that participants develop a belief in the context of the experiment. As described in more detail later, Experiment 4 assessed this possibility using prestudy JOLs (e.g., Mueller, Tauber, & Dunlosky, 2013; Mueller et al., 2014). Experiment 1 Experiment 1 examined whether differences in effective study time account for the effect of motoric fluency on JOLs by implementing a fixed 4 s of study time after participants copied down a word. If the effect really is due to motoric fluency, it should persist under these conditions; if the effect is due to differences in effective study time, the current implementation should eliminate it. Method Participants Twenty-four undergraduate students from the University of North Carolina at Chapel Hill participated in exchange for course

118

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

Table 1 Mean (SE) median writing times (for Experiments 1, 2, and 4) or processing times (for Experiments 3a and 3b) in seconds

Experiment Experiment Experiment Experiment Experiment

1 2 3a 3b 4

Dominant

Non-dominant

6.14 5.87 3.21 3.96 5.15

9.87 8.83 3.24 3.74 8.13

(0.38) (0.34) (0.26) (0.45) (0.32)

(0.64) (0.34) (0.23) (0.41) (0.32)

credit. No additional demographic information was collected. The participants were naïve to the purpose of the experiment and had not participated in any related study. Materials and design The critical words consisted of 40 four-letter words, half high frequency (100–500 per million) and half low frequency (1–22; Kucera & Francis, 1967).1 Four additional words were presented at the beginning and end of the list as primacy and recency buffers, and two more were used as practice. Writing hand and word frequency were factorially manipulated within subjects. The study list was pseudo-randomly ordered such that no more than two trials of a hand condition were presented consecutively. Two versions of the list were constructed, counterbalancing words across hand. Procedure The experiment consisted of three phases: study, distraction, and test, and replicated the main design used by Susser and Mulligan (2015). During the study phase, participants viewed individual words in the center of a computer screen and were told to study them for later. The words Right Hand or Left Hand were displayed below each word on the right or left side of the screen, respectively, indicating the hand that participants should use to copy the word. Participants were instructed to print each word with the appropriate hand on its own note card. They were asked to write as quickly and accurately as possible and to then press the Space Bar; this recorded the writing time for that word. The word remained on the screen for another 4 s after the Space Bar was pressed, and then participants rated their confidence in remembering it from 0 (not at all confident) to 100 (extremely confident) and pressed Enter to proceed to the next trial. In the distraction phase, participants completed arithmetic problems for three minutes. Then, in the test phase, participants were given a 5-min free recall test. Finally, participants filled out the Edinburgh Handedness Inventory (Oldfield, 1971) to assess handedness. Consistent with standard scoring, participants who scored above 0 were considered right handed, and participants who scored below 0 were considered left handed. This outcome was used to code the words as written with either the dominant or non-dominant hand. Results and discussion Twenty-one of the 24 participants were right handed. To examine writing times during the study phase, a paired-samples t-test was conducted on median writing times (Table 1). Not surprisingly, this analysis revealed a significant 1 Although not a focus of the present study, high- and low-frequency words were included in the original Susser and Mulligan (2015) study because word frequency has sometimes, but not always, affected JOLs (e.g., Benjamin, 2003; cf. Tullis & Benjamin, 2012). In that paper, word frequency had no significant effect on JOLs or on recall (see Susser & Mulligan, 2015, for details). To maintain comparability with Susser and Mulligan, the same materials were used in this and subsequent experiments; however, word frequency continued to have no reliable effect on JOLs or recall, so we collapsed across this variable in the reported results.

effect, t(23) = 9.55, p < 0.001, d = 1.95, with longer writing times for the non-dominant than dominant hand. A paired-samples t-test conducted on JOLs revealed that they were higher for items written with the dominant (M = 46.89, SD = 17.48) than non-dominant hand (M = 42.84, SD = 18.74), t(23) = 2.42, p = 0.024, d = 0.49. The same analysis revealed no difference in recall for the dominant (M = 0.17, SD = 0.11) and non-dominant hand (M = 0.18, SD = 0.11), t(23) = 0.34, p = 0.738, d = 0.07. The relative accuracy of JOLs as indexed by gamma correlations (Nelson, 1984) was significantly greater than 0, G = 0.31, SE = 0.06, t(22) = 4.96, p < 0.001, d = 1.03.2 The present results replicated those of Susser and Mulligan (2015) in showing that hand dominance (1) affected motoric fluency (as indexed by writing times), (2) affected JOLs, and (3) did not affect recall. Critically, the effect of motoric fluency on JOLs persisted even when post-writing study time was held constant for all items, suggesting that differences in effective study time in Susser and Mulligan’s original design did not drive the JOL effect. Experiment 2 The findings of Experiment 1 argue that differences in effective study time do not account for the motoric fluency effect on JOLs: the effect occurs whether one controls the nominal study time (as in the original experiments; Susser & Mulligan, 2015) or the effective, post-writing study time (as in Experiment 1). Experiment 2 further analyzes the locus of the effect by examining whether motoric fluency itself is really its basis. As noted in the Introduction, writers not only experience the motoric feeling of writing but are also exposed to the visual result of the writing process, raising the possibility that the current effect is due not to motoric fluency but to perceptual feedback. Writing with the non-dominant hand both feels less motorically fluent and produces visual text that is less legible and clear than writing with the dominant hand, which may reduce confidence (in the present case, memorial confidence) in that information (Briñol & Petty, 2003). This possibility comports with the more general finding that manipulations of the visual (or auditory) properties of verbal material often impact metamemory (Besken & Mulligan, 2013, 2014; McDonough & Gallo, 2012; Rhodes & Castel, 2008, 2009; Susser, Mulligan, & Besken, 2013). The following experiments tease apart the contributions of motor fluency and perceptual feedback. We first examine the effect of motoric fluency in the absence of visual feedback. To achieve this, we had participants copy the study words on carbon paper using a stylus. This eliminated the visual product of the handwriting but retained its motoric feeling. Method Participants Eighteen undergraduate students from the University of North Carolina at Chapel Hill participated in exchange for course credit. No additional demographic information was collected. The participants were naïve to the purpose of the experiment and had not participated in any related study. Materials and design The materials and design were identical to Experiment 1. 2 One participant gave a JOL of 50 for every item; therefore, gamma could not be computed. Providing the same JOL on all trials may cause one to question whether the participant was following directions. On the side of caution, all analyses were repeated without this participant. The pattern of results did not change.

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

Procedure The instructions for the study phase were identical to Experiment 1 with the exception that participants were told that they would be writing on carbon paper and would not see their own handwriting. Participants were shown the stylus and given a demonstration of how the carbon paper worked to ensure that they would try to write accurately. The study trials proceeded as in Experiment 1 with a few modifications. First, rather than writing on note cards, participants used a clipboard with three sheets of paper. The top sheet was numbered for each word in the experiment, the middle sheet was a piece of carbon paper, and the bottom sheet was a blank piece of paper. Participants were told to write as they normally would and that they did not need to use any extra pressure. Second, each study word appeared on the computer screen with its number to help participants follow along on their sheet of paper. Participants were instructed to print each word with the appropriate hand in the space provided on the sheet, as quickly and accurately as possible. After writing the word, participants pressed the Space Bar to measure writing time; however, the word remained on the screen for a fixed total of 13 s (consistent with the original procedure of Susser and Mulligan (2015)). Participants then rated their confidence in remembering the word. The rest of the experiment was identical to Experiment 1.

119

experiment. Alternatively, Briñol and Petty (2003) argued that the visual feedback of hand writing plays a critical role in people’s metacognitive judgments. Because this experiment might be implemented in a number of ways, we conducted it twice to ensure generalizability over some of the methodological details. In Experiment 3a, the written products of two subjects from a prior experiment (one for each counterbalancing condition) were used, and participants were asked to verify that the written word on the note card matched the study word presented on the screen. This selection of materials guaranteed that participants were exposed to the same (legible) materials and that each word was clearly produced by a dominant or a nondominant writing hand, but it limited the variability present in the actual writing samples. In Experiment 3b, we yoked each participant to a prior student’s note cards from an earlier experiment in which the motoric fluency effect on JOLs was found. This ensured that the variability in visual feedback present in an experiment that produced the effect would also be present in the current experiment. Furthermore, to ensure that participants were actually engaged in processing the visual appearance of the written information on the cards, we had them sort the cards by hand dominance.5 Experiment 3a method

Results and discussion Sixteen of the 18 participants were right handed. A paired-samples t-test conducted on median writing times3 (Table 1) revealed that writing took longer for non-dominant than dominant items, t(17) = 12.56, p < 0.001, d = 2.96.4 A paired-samples t-test conducted on JOLs also revealed a significant effect, t(17) = 5.64, p < 0.001, d = 1.33, with higher JOLs for the dominant (M = 46.22, SD = 20.03) than non-dominant (M = 39.63, SD = 19.69) items. The same analysis conducted on recall performance revealed no significant difference in memory for the dominant (M = 0.21, SD = 0.13) and non-dominant (M = 0.18, SD = 0.12) items, t(17) = 0.99, p = 0.338, d = 0.23. The relative accuracy of JOLs was significantly greater than 0, G = 0.35, SE = 0.10, t(17) = 3.57, p = 0.002, d = 0.84. The results of Experiment 2 are consistent with those of the prior experiment and prior research (Susser & Mulligan, 2015) in showing that hand dominance affects metamemory. However, in the present case we obtained this result without the potential influence of visual feedback. It appears that the motoric component of writing can affect JOLs even when it is the only cue available. This finding indicates that visual feedback is not necessary for the effect and that motoric fluency alone is sufficient. Experiments 3a and 3b Experiment 2 demonstrates that the experience of motoric fluency without visual feedback can produce an effect on JOLs. Experiments 3a and 3b examined the complementary circumstance of visual feedback without motor experience. In particular, participants received the product of writing from another student’s dominant and non-dominant hands without, of course, having the accompanying motor experience. If the motoric fluency effect is purely motoric, then the effect on JOLs should be eliminated in this 3 Writing times were excluded on 3.9% of the trials because participants either failed to press the Space Bar after writing a word or did not complete the word within the allotted time. 4 This analysis revealed an effect involving word frequency – a significant main effect, F(1, 17) = 7.65, p = 0.013, MSE = 1.68  105, g2p = 0.31, showing quicker writing times for high- than low-frequency items.

Participants Thirty-two undergraduate students from the University of North Carolina at Chapel Hill participated in exchange for course credit. No additional demographic information was collected. The participants were naïve to the purpose of the experiment and had not participated in any related study. Materials and design The materials and design were similar to the prior experiments. For this experiment, the researchers selected two sets of note cards written by students in earlier versions of the experiment, counterbalancing words across hand condition. The note cards were chosen based on their legibility but also on the obviousness of the hand condition (a research assistant successfully categorized all words as having been written with the writer’s dominant or nondominant hand). Procedure Each study trial consisted of a word presented on the computer screen, accompanied with the Right Hand or Left Hand instruction. Participants were told that students in a previous experiment printed these words on note cards using the hand indicated on the trial, and that, now, they would view the writing of one right-handed student. Participants were instructed to read the word on the screen, turn over the top note card, silently read the word on the card, and verify that it matched the one on the screen (it always did). After participants read and verified the word, they pressed the Space Bar, recording the processing time for that word. The word remained on the screen for a total of 13 s, and then participants made their JOL. The distraction and test phases were identical to the previous experiments. Experiment 3b method Participants Thirty-two undergraduate students from the University of North Carolina at Chapel Hill participated in exchange for course 5 Throughout the course of this experiment, researchers also monitored participants to make sure they effectively followed task instructions and were sorting the cards.

120

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

credit. No additional demographic information was collected. The participants were naïve to the purpose of the experiment and had not participated in any related study. Materials, design, and procedure The methods were identical to Experiment 3a with the following alterations. The main alteration was that in Experiment 3b each participant was yoked to a participant from an earlier experiment that found the motoric fluency effect (Susser & Mulligan, 2015, Experiment 1) and presented with the set of cards written by that participant. During the study phase, participants viewed words on a computer screen without the hand condition listed. Participants were told that students in a previous experiment printed these words on note cards with their dominant and non-dominant hands, and that, now, they would view the writing of one of these students. Participants were instructed to read the word on the screen, turn over the top note card, silently read the word on the card, verify that it matched the one on the screen (it always did), and then sort the card as being written with the dominant or non-dominant hand. After sorting, participants pressed the Space Bar, and after the full 13 s, they made their JOL. The distraction and test phases were identical to the previous experiments. Results Experiment 3a A paired-samples t-test revealed no effect of hand on median processing times6 (Table 1), t(29) = 0.50, p = 0.620, d = 0.09, or on JOLs, t(31) = 0.23, p = 0.817, d = 0.04, which were similar for the dominant (M = 51.28, SD = 22.84) and non-dominant (M = 51.55, SD = 22.86) words. Likewise, this analysis found no difference in recall between the dominant (M = 0.23, SD = 0.12) and non-dominant (M = 0.21, SD = 0.12) words, t(31) = 1.38, p = 0.177, d = 0.24. Relative accuracy was significantly greater than 0, G = 0.30, SE = 0.06, t(29) = 4.74, p < 0.001, d = 0.86.7 The results of Experiment 3a provide preliminary evidence that perceptual feedback does not contribute to the motoric fluency effect: when participants received the visual feedback of the writing of other students without the accompanying motor experience, their JOLs did not differ across hand condition. Because this is a null result, we conducted a power analysis using the average effect size of the present Experiments 1 and 2 and Experiment 2 from Susser and Mulligan (2015) (d = 1.19), experiments that were otherwise similar to the current experiment. Power exceeded 0.99 to detect an effect of this size and exceeded 0.89 to detect an effect half the size; therefore, the present experiment had substantial power. Experiment 3b Participants correctly sorted 97.4% of the words, supporting the idea that they paid attention to the written information on the cards. A paired-samples t-test run on median processing times8 (Table 1) revealed a significant effect of hand, t(29) = 2.45, p = 0.020, d = 0.45. Participants sorted the non-dominant words more quickly than the dominant words. 6 Two participants failed to press the Space Bar on the vast majority of trials (>70%); their processing time data were excluded. An additional 3.0% of the processing times were excluded for the same reason. 7 One participant gave a JOL of 50 for every item and another gave a JOL of 10 for every item; therefore, gamma could not be computed. Removing these participants from all analyses did not change the pattern of results. 8 Two participants failed to press the Space Bar on the vast majority of trials (>70%). Their processing time data were excluded. Less than 1% of the additional processing times were excluded for the same reason.

The same paired-samples t-test revealed no significant difference in JOLs for dominant (M = 52.68, SD = 17.54) and nondominant (M = 50.40, SD = 16.61) words, t(31) = 1.84, p = 0.075, d = 0.33, and no effect on recall performance, t(31) = 0.17, p = 0.867, d = 0.03. Participants recalled a similar proportion of non-dominant (M = 0.19, SD = 0.10) and dominant (M = 0.19, SD = 0.14) items. The relative accuracy of JOLs was significantly greater than 0, G = 0.24, SE = 0.06, t(30) = 4.04, p < .001, d = 0.73.9 Consistent with Experiment 3a, participants in Experiment 3b did not appear to use perceptual feedback as a cue for their JOLs, although there was a trend toward an effect in this experiment. As in Experiment 3a, we conducted a power analysis using the average effect size of Experiments 1 and 2 and Susser and Mulligan’s (2015) Experiment 2. Power exceeded 0.99 to detect an effect of this size and exceeded 0.89 to detect an effect one half the size. Combined analyses To provide an even more powerful test of whether visual feedback alone produces an impact on JOLs, we ran a paired-samples ttest on the combined data from Experiments 3a and 3b. We found no effect of hand, t(63) = 1.18, p = 0.243, d = 0.15. The power to detect an effect one half the size described above exceeded 0.99. Thus, we had overwhelming power to find an effect of visual feedback on JOLs of the size revealed when motoric fluency is actually experienced, and one much smaller. Additionally, we assessed whether the effect of hand on JOLs significantly changed when moving from the motor-only paradigm of Experiment 2 to the visual feedback paradigms of Experiments 3a and 3b. We approached this assessment in three different ways. First, we conducted separate 2 (Experiment)  2 (Hand) mixed analyses of variance (ANOVAs) using Experiment 3a’s and 3b’s individual data versus the data from Experiment 2. Then we ran the same mixed ANOVA using the combined data from Experiments 3a and 3b. In all three analyses the critical interaction between experiment and hand was significant. Comparing Experiment 3a with Experiment 2, we observed a significant main effect of hand, F(1, 48) = 12.76, p = 0.001, MSE = 18.09, g2p = 0.21, no main effect of experiment, F(1, 48) = 1.77, p = 0.189, and a significant interaction, F(1, 48) = 15.02, p < .001, MSE = 18.09, g2p = 0.24. Comparing Experiment 3b with Experiment 2, we found a significant main effect of hand, F(1, 48) = 22.39, p < 0.001, MSE = 20.30, g2p = 0.32, no main effect of experiment, F(1, 48) = 2.69, p = 0.107, and a significant interaction, F(1, 48) = 5.28, p = 0.026, MSE = 20.30, g2p = 0.10. In the combined experiment analysis, we found a significant main effect of hand, F(1, 80) = 19.29, p < .001, MSE = 21.07, g2p = 0.19; no main effect of experiment, F(1, 80) = 2.64, p = 0.108, and a significant interaction, F(1, 80) = 10.42, p = 0.002, MSE = 21.07, g2p = 0.12. The significant interaction in the three analyses reveals that the effect of hand on JOLs was larger in Experiment 2, when only the motoric component was present, than in Experiments 3a and 3b, when only visual information was present. Jointly, this set of analyses indicates that the motoric fluency effect is significantly larger when motoric fluency is experienced than when it is not (and only visual feedback is provided), and that when visual feedback is experienced without motoric fluency, there is little consistent evidence for the effect. Discussion The results of Experiments 3a and 3b indicate that visual information from words written with a dominant versus non-dominant

9 One participant gave a JOL of 60 for every item; therefore, gamma could not be computed. Removing this participant did not change the pattern of results.

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

hand does not strongly contribute to the motoric fluency effect. This outcome is inconsistent with Briñol and Petty’s (2003) proposal that the perceptual characteristics of non-dominant writing may affect metacognitive judgments, but is consistent with the notion that the present effect requires motor experience. The result was consistent across two versions of the experiment using different sets of writing samples and different procedures for how the participants interacted with the hand writing. It should be noted that the current experiments were not designed to assess the general proposition that variation in the perceptual characteristics of words can influence JOLs. As discussed earlier, it is already known that such variations can affect metamemory, at least under some circumstances (e.g., Besken & Mulligan, 2013; McDonough & Gallo, 2012; Rhodes & Castel, 2008). The question for the current set of experiments was whether a particular perceptual variation (words written with the dominant or non-dominant hand) influenced JOLs in the context of our motoric fluency experimental design. A priori, this is a plausible possibility, but clearly one that does not actually apply in the present case: little effect on JOLs emerged despite substantial power. Given that perceptual variation in words can influence metamemory, why does variation in handwriting not contribute here? A couple of possibilities exist. It may be that reading words in different hand writing does not induce a sufficient change in subjective fluency or belief state to influence JOLs. Alternatively, it could be because participants see every study word in the same fluent manner when the words are first presented on the computer screen before seeing the handwritten version. Perhaps this initial processing overrides later variation in perceptual processing. These possibilities are certainly amenable to future exploration, but for present purposes deciding between them is unimportant. What is important is that in the paradigm that produces the motoric fluency effect, processing of the visual characteristics of the hand writing does not appear to substantially contribute. Experiment 4 The results of Experiments 1–3b suggest that the motor component is a cue for JOLs and is necessary to produce the motoric fluency effect. An additional issue, though, is whether the effect stems from experienced fluency or beliefs about the manipulation (e.g., Koriat et al., 2004; Mueller et al., 2014). Susser and Mulligan’s (2015) initial demonstration suggested that actual fluency was driving JOLs: writing times mediated the effect of hand on JOLs, and participants did not report a priori beliefs about an effect of writing hand on memory. Fluency and a priori beliefs can be thought to represent endpoints on a continuum; however, it is also possible that participants develop beliefs in the context of an experiment. In other words, experience with the materials and manipulation may be needed, but this experience informs beliefs that people form on-line. One way to tap this type of belief is with prestudy JOLs (e.g., Castel, 2008; Mueller et al., 2013), in which participants are informed of the status of the upcoming item (i.e., what condition it is in) but make a JOL before actually experiencing it. If an effect on prestudy JOLs is found, it suggests that on-line beliefs and not just processing fluency contribute to the effect. Alternatively, if a manipulation affects typical post-study JOLs but not prestudy JOLs, this is evidence against a belief-based contribution. Method Participants Twenty-four undergraduate students from the University of North Carolina at Chapel Hill participated in exchange for course

121

credit. No additional demographic information was collected. The participants were naïve to the purpose of the experiment and had not participated in any related study. Materials and design The materials and design were identical to Experiment 1. Procedure The experiment replicated the motoric fluency experiment of Susser and Mulligan (2015, Experiment 2) but used self-paced prestudy JOLs occurring before each study item in place of traditional post-study JOLs. Specifically, before each item, participants saw, ‘‘You are about to study a [Right Hand]/[Left Hand] word. Please rate your memory confidence (0–100).” Participants typed in their confidence from 0 (not at all confident) to 100 (extremely confident) for the upcoming (but not yet seen) word. Participants then viewed and copied down the word on a note card, pressing the Space Bar when done writing. The word remained on the screen for a total of 13 s. The rest of the experiment was identical to the prior ones. Results and discussion Twenty-three of the 24 participants were right handed. A paired-samples t-test conducted on median writing times10 (Table 1) showed that participants took longer to write the nondominant than dominant items, t(21) = 13.52, p < 0.001, d = 2.88.11 Prestudy JOLs were higher for the dominant (M = 53.15, SD = 26.20) than non-dominant (M = 48.27, SD = 24.13) items, t (23) = 2.73, p = 0.012, d = 0.56, indicating that the motoric fluency effect occurs even with JOLs made before experiencing specific items. To follow up on this analysis, we assessed whether the effect of hand on JOLs changed across the study list. If prestudy JOLs index beliefs developed in the context of the experiment, two possibilities arise. First, the effect may slowly develop over the course of the experiment such that the effect of hand is minimal initially and then increases as participants gain more experience writing with their two hands. Alternatively, the belief may develop quickly. The motor experience and feeling of disfluency when writing with the non-dominant hand is quite marked and may be salient even with little practice, in which case the motoric fluency effect may well be substantial early in the experiment and not increase in size. To assess these possibilities, we examined the effect of hand across the first and second halves of the study list (Table 2). We ran a 2 (Hand: Dominant vs. Non-dominant)  2 (List half: First vs. Second) repeated-measures ANOVA. The analysis revealed a significant main effect of hand, F(1, 23) = 7.41, p = 0.012, MSE = 76.63, g2p = 0.24, and a significant main effect of list half, F(1, 23) = 16.18, p = 0.001, MSE = 67.31, g2p = 0.41, with JOLs decreasing from the first half to the second half of the list. However, there was no significant interaction, F(1, 23) = 0.96, p = 0.337, indicating that the effect of hand did not change over time.12 Recall performance did not differ between the dominant (M = 0.18, SD = 0.12) and non-dominant (M = 0.19, SD = 0.15) items, t(23) = 0.46, p = 0.652, d = 0.10. The relative accuracy of JOLs did

10 Two participants failed to press the Space Bar on the vast majority of trials (>70%); their writing time data were excluded. An additional 4.2% of the data were excluded for the same reason. 11 This analysis found an effect involving word frequency – a significant interaction between it and hand, F(1, 21) = 8.31, p = 0.009, MSE = 1.00  105, g2p = 0.28. For the dominant hand, low-frequency words tended to be written more slowly than highfrequency words. For the non-dominant hand, high-frequency words tended to be written more slowly. 12 We also ran this analysis with the buffer items included and using list quarters both with and without buffer items. The pattern remained the same across all analyses.

122

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

Table 2 Mean (SE) JOLs across hand and list half in Experiment 4

First half of study list Second half of study list

Dominant hand

Non-dominant hand

56.92 (4.95) 50.05 (5.89)

51.92 (4.72) 44.65 (5.23)

not differ from 0, G = 0.07, SE = 0.09, t(20) = 0.76, p = 0.459, d = 0.16, which is unsurprising given that JOLs were made prior to actually seeing the study items.13 The results of Experiment 4 provide further insight into the manner in which the motoric experience of writing affects JOLs. Susser and Mulligan (2015) showed evidence that the experience of writing fluency contributes to the effect. Here, with prestudy JOLs, we have shown that people may additionally generate a belief about the effect of hand in the context of the experiment. The fact that the effect did not increase over the halves of the study list may seem inconsistent with a developed belief; it would be plausible to think that participants would initially have little or no belief (consistent with Susser and Mulligan’s questionnaire results) but then slowly generate one as they interact with the task. However, this does not seem to be the case. Instead, participants may simply need some experience writing with each hand to develop (or trigger) a belief relatively quickly. This notion is plausible given how marked and salient the hand manipulation is. Therefore, Experiment 4 implies that, in the context of this experimental design, a type of belief can contribute to the motoric fluency effect. We return to this point in the General Discussion.

General discussion One aim of metamemory research is to identify the cues that people use to make predictions about their future memory. Susser and Mulligan (2015) suggested that motoric fluency might be one of these cues, and the current experiments both corroborate and complicate this notion. The first two experiments showed that JOLs are higher for words written with the dominant than nondominant hand even when effective (post-writing) study time is controlled (Experiment 1) and when no visual feedback is presented (Experiment 2). The second two experiments (Experiments 3a and 3b) showed that JOLs are not significantly affected when only visual feedback is available (without motor experience). The final experiment (Experiment 4) showed that the motoric fluency effect occurs even with prestudy JOLs, indicating that although the effect requires the motoric experience of writing with the dominant and non-dominant hands, it may be influenced by beliefs informed by that experience. A substantial body of research has investigated the extent to which manipulations of fluency affect metacognition, and metamemory in particular. Claims have been made that the ease of both perceptual and conceptual processing can influence predictions of memorability even when such manipulations fail to affect memory (e.g., Rhodes and Castel, 2008, 2009) or do so in the opposite direction (e.g., Besken and Mulligan, 2013, 2014). The motoric fluency effect appears to fit neatly into this tradition, extending it to the realm of motor as opposed to perceptual or conceptual fluency (Susser and Mulligan, 2015). However, in research on fluency and metamemory, there is debate as to the actual bases of the effects (e.g., Mueller et al., 2013; Susser et al., 2016; Undorf and Erdfelder, 2015), and effects initially attributed to fluency are 13 One participant failed to recall any items, a second participant gave a JOL of 90 for every item, and a third participant gave a JOL of 100 for every item. Gamma could not be computed for these participants. Removing the latter two participants did not change the pattern of results.

sometimes found to have other bases (Mueller et al., 2014; cf. Rhodes and Castel, 2008). The same challenges confront the notion that motoric fluency influences metamemory. Does the motoric fluency effect actually depend on the motor component of the manipulation? The first question regarding such effects is whether the manipulation actually affects the ease of processing in the appropriate modality (e.g., Besken and Mulligan, 2014; Mueller et al., 2014). The current manipulation of writing hand passes this test, perhaps not surprisingly, as it had a substantial effect on an objective measure of the motor behavior (writing time) in all of the present experiments and those of Susser and Mulligan (2015). However, this by itself does not ensure that the motor component actually produces the effect. The present experiments argue against two reasonable alternative possibilities. First, in Susser and Mulligan’s original experiments, the effective, postwriting study time was greater in the dominant than nondominant condition. Given that JOLs are sensitive to study time (Koriat and Ma’ayan, 2005), this difference, rather than difference in motoric fluency, might have produced the effect. Experiment 1 ruled out this possibility. Second, hand writing may influence judgments through the visual appearance of the text, an apparent effect of motoric fluency that is actually due to perceptual feedback (Briñol and Petty, 2003). Contradicting this explanation, Experiment 2 showed that visual feedback is not necessary to produce the effect, and Experiments 3a and 3b indicated that it is not sufficient. One might wonder whether visual feedback still plays a role in metamemory in conjunction with motoric fluency. Perhaps the visual appearance of the writing is only effective if it is the product of one’s own writing. Our results show no evidence for this possibility. If visual feedback contributes to the motoric fluency effect, the effect should be diminished when this feedback is eliminated, as in the present Experiment 2. However, the effect size of hand in this experiment – when the motor component was the only cue available – is the largest of the three experiments finding the motoric fluency effect in this paper (Experiments 1, 2, and 4). Consequently, there is no evidence that removing visual feedback diminishes the motoric fluency effect, and any trend is in the opposite direction. Experiment 4 speaks to the distinction between experienceand belief-based influences in metamemory. The experiencebased influence is rooted in non-analytic subjective experiences with the stimulus materials, such as the feelings of fluency that are often thought to impact metacognitive judgments (e.g., Rhodes and Castel, 2008). In contrast, beliefs refer to theories or hypotheses about how memory operates. These theories are then applied in a particular experimental context to yield a metamemorial judgment. Beliefs might exist a priori or be produced as participants interact with the experimental materials and task. These influences are not mutually exclusive, though; it may well be that a metamemorial cue has an effect on one or multiple of these bases (see Mueller et al., 2016, for a discussion). As noted earlier, fluency effects in metamemory are often attributed to experience, such that the experienced ease or difficulty of processing a particular stimulus is thought to influence the JOL for that stimulus. One way to assess such a possibility is with the partial correlation analysis suggested by Mueller et al. (2014). First, the effect of the purported fluency variable (in this case, writing hand) is assessed on a relevant measure of fluency (writing time) to verify that the manipulation actually affects fluency at all. Once this is established, a partial correlation analysis is conducted to determine if the relationship between the fluency manipulation (writing hand) and metamemory is mediated by the actual change in fluency produced by the manipulation (writing time). In the case of the motoric fluency effect, Susser and Mulligan (2015) found that writing hand, of course,

J.A. Susser et al. / Journal of Memory and Language 95 (2017) 116–123

affects writing time, and that the correlation between writing hand and JOLs was significantly diminished when writing time was controlled, consistent with an experience-based contribution to the effect. To assess a priori beliefs, it is standard to use a questionnaire in which participants read a description of an experiment and predict recall performance for different conditions without actually having the experience (e.g., of varying degrees of perceptual or motoric fluency) (see Koriat et al., 2004; Mueller et al., 2014). When this method was applied to motoric fluency, participants predicted no difference in memory for words written with the dominant versus non-dominant hand, indicating no a priori belief regarding this metamemorial cue (Susser and Mulligan, 2015, Experiment 3). Finally, the use of prestudy JOLs provides the opportunity to tap beliefs in a different way, when participants are engaged with the experimental procedure. Technically, prestudy JOLs are capable of revealing either a priori or developed beliefs. Although Susser and Mulligan (2015), using a questionnaire, found no support for a priori beliefs as a basis of the motoric fluency effect, the results of the present Experiment 4 cloud this conclusion somewhat. Prestudy JOLs are most diagnostic of beliefs developed on-line during the experiment, and Experiment 4 showed that the motoric fluency effect emerged with these JOLs. However, the effect did not grow over time, as might be expected if it truly developed over the course of the experiment. Instead, it may be that the manipulation is so marked that a belief develops rather quickly. It is also clear that the questionnaire and prestudy JOL approaches differ in a number of ways. First, the prestudy JOL paradigm resembles a typical JOL experiment in which JOLs are calculated using a number of items. In a questionnaire design, participants often provide a single ‘‘JOL” for the relevant classes of items. One might think that this difference could contribute to the disparity in results across Experiment 4 and Susser and Mulligan’s (2015) Experiment 3. However, it should be noted that the list-end JOL method also elicits a single JOL for each class of items, and this method produced a significant motoric fluency effect on JOLs (Susser & Mulligan, Experiment 1). So, eliciting a single JOL for each stimulus type is not by itself a barrier to observing the motoric fluency effect. Second, the nature of the study itself greatly differs across the two designs. A questionnaire can be completed in a few minutes, and participants do not encounter the manipulation first hand. A prestudy JOL design can be more engaging for participants, and they experience the manipulation. Therefore, it is possible that the two paradigms invoke in participants different approaches to the task (see Mueller et al., 2016). It is beyond the scope of this paper to further compare the two paradigms, and additional research ought to do so, but for the present purposes, our findings imply that a form of belief contributes to the motoric fluency effect. Therefore, based on standard assessments of belief and experience, there is evidence that writing hand influences metamemory through experience proper (Susser and Mulligan, 2015) and through experience-derived belief. In conclusion, the present results imply that the motoric fluency effect is appropriately named and relies on the motoric experience of writing. This finding complements the majority of earlier JOL work and demonstrates that the fluency of motor actions – in addition to perceptual or conceptual fluency – can affect people’s memorial confidence in that information, even when it does not affect actual memory performance. Acknowledgments This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

123

References Alban, M. W., & Kelley, C. M. (2013). Embodiment meets metamemory: Weight as a cue for metacognitive judgments. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39, 1628–1634. http://dx.doi.org/10.1037/a0032420. Alter, A. L., & Oppenheimer, D. M. (2009). Uniting the tribes of fluency to form a metacognitive nation. Personality and Social Psychology Review, 13, 219–235. http://dx.doi.org/10.1177/1088868309341564. Benjamin, A. S. (2003). Predicting and postdicting the effects of word frequency on memory. Memory & Cognition, 31, 297–305. http://dx.doi.org/10.3758/BF03194388. Benjamin, A. S., Bjork, R. A., & Schwartz, B. L. (1998). The mismeasure of memory: When retrieval fluency is misleading as a metamnemonic index. Journal of Experimental Psychology: General, 127, 55–68. http://dx.doi.org/10.1037/00963445.127.1.55. Besken, M., & Mulligan, N. W. (2013). Easily perceived, easily remembered? Perceptual interference produces a double dissociation between metamemory and memory performance. Memory & Cognition, 41, 897–903. http://dx.doi.org/ 10.3758/s13421-013-0307-8. Besken, M., & Mulligan, N. W. (2014). Perceptual fluency, auditory generation, and metamemory: Analyzing the perceptual fluency hypothesis in the auditory modality. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 429–440. http://dx.doi.org/10.1037/a0034407. Briñol, P., & Petty, R. E. (2003). Overt head movements and persuasion: A selfvalidation analysis. Journal of Personality and Social Psychology, 84, 1123–1139. http://dx.doi.org/10.1037/0022-3514.84.6.1123. Castel, A. D. (2008). Metacognition and learning about primacy and recency effects in free recall: The utilization of intrinsic and extrinsic cues when making judgments of learning. Memory & Cognition, 36, 429–437. http://dx.doi.org/ 10.3758/MC.36.2.429. Koriat, A., Bjork, R. A., Sheffer, L., & Bar, S. K. (2004). Predicting one’s own forgetting: The role of experience-based and theory-based processes. Journal of Experimental Psychology: General, 133, 643–656. http://dx.doi.org/10.1037/ 0096-3445.133.4.643. Koriat, A., & Ma’ayan, H. (2005). The effects of encoding fluency and retrieval fluency on judgments of learning. Journal of Memory and Language, 52, 478–492. http://dx.doi.org/10.1016/j.jml.2005.01.001. Kucera, H., & Francis, W. N. (1967). Computational analysis of present-day American English. Providence, RI: Brown University Press. Matvey, G., Dunlosky, J., & Guttentag, R. (2001). Fluency of retrieval at study affects judgments of learning (JOLs): An analytic or nonanalytic basis for JOLs? Memory & Cognition, 29, 222–233. http://dx.doi.org/10.3758/BF03194916. McDonough, I. M., & Gallo, D. A. (2012). Illusory expectations can affect retrievalmonitoring accuracy. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38(2), 391–404. http://dx.doi.org/10.1037/a0025548. Mueller, M. L., Dunlosky, J., & Tauber, S. K. (2016). The effect of identical word pairs on people’s metamemory judgments: What are the contributions of processing fluency and beliefs about memory? The Quarterly Journal of Experimental Psychology, 69(4), 781–799. http://dx.doi.org/10.1080/17470218.2015.1058404. Mueller, M. L., Dunlosky, J., Tauber, S. K., & Rhodes, M. G. (2014). The font-size effect on judgments of learning: Does it exemplify fluency effects or reflect people’s beliefs about memory? Journal of Memory and Language, 70, 1–12. http://dx.doi. org/10.1016/j.jml.2013.09.007. Mueller, M. L., Tauber, S. K., & Dunlosky, J. (2013). Contributions of beliefs and processing fluency to the effect of relatedness on judgments of learning. Psychonomic Bulletin & Review, 20, 378–384. http://dx.doi.org/10.3758/s13423012-0343-6. Nelson, T. O. (1984). A comparison of current measures of the accuracy of feelingof-knowing predictions. Psychological Bulletin, 95, 109–133. http://dx.doi.org/ 10.1037/0033-2909.95.1.109. Oldfield, R. C. (1971). The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia, 9, 97–113. http://dx.doi.org/10.1016/0028-3932 (71)90067-4. Rhodes, M. G., & Castel, A. D. (2008). Memory predictions are influenced by perceptual information: Evidence for metacognitive illusions. Journal of Experimental Psychology: General, 137, 615–625. http://dx.doi.org/10.1037/a0013684. Rhodes, M. G., & Castel, A. D. (2009). Metacognitive illusions for auditory information: Effects on monitoring and control. Psychonomic Bulletin & Review, 16, 550–554. http://dx.doi.org/10.3758/PBR.16.3.550. Susser, J. A., Jin, A., & Mulligan, N. W. (2016). Identity priming consistently affects perceptual fluency but only affects metamemory when primes are obvious. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42(4), 657–662. http://dx.doi.org/10.1037/xlm0000189. Susser, J. A., & Mulligan, N. W. (2015). The effect of motoric fluency on metamemory. Psychonomic Bulletin & Review, 22, 1014–1019. http://dx.doi. org/10.3758/s13423-014-0768-1. Susser, J. A., Mulligan, N. W., & Besken, M. (2013). The effects of perceptual fluency and list composition on judgments of learning (JOLs). Memory & Cognition, 41, 1000–1011. http://dx.doi.org/10.3758/s13421-013-0323-8. Tullis, J. G., & Benjamin, A. S. (2012). The effectiveness of updating metacognitive knowledge in the elderly: Evidence from metamnemonic judgments of word frequency. Psychology and Aging, 27, 683–690. http://dx.doi.org/10.1037/ a0025838. Undorf, M., & Erdfelder, E. (2015). The relatedness effect on judgments of learning: A closer look at the contribution of processing fluency. Memory & Cognition, 43, 647–658. http://dx.doi.org/10.3758/s13421-014-0479-x.