Cross-cultural variability of component processes in autobiographical remembering: Japan, Turkey, and the U.S.A.

Abstract

Although the underlying mechanics of autobiographical memory may be identical across cultures, the processing of information differs. Undergraduates from Japan, Turkey, and the U.S.A. rated 30 autobiographical memories on 15 phenomenological and cognitive properties. Mean values were similar across culture, with means from the Japanese sample being lower on most measures but higher on belief in the accuracy of their memories. Correlations within individuals were also similar across cultures, with correlations from the Turkish sample being higher between measures of language and measures of recollection and belief. For all three cultures, in multiple regression analyses, measures of recollection were predicted by visual imagery, auditory imagery, and emotions, whereas measures of belief were predicted by knowledge of the setting. These results show subtle cultural differences in the experience of remembering.

In psychology, the cross-cultural study of autobiographical memory has focused on two areas: the development of narrative in autobiographical memory (Han, Leichtman, & Wang, 1998; Jin & Naka, 2002; Mullen & Yi, 1995; Wang & Brockmeister, 2002) and the distribution of autobiographical memories across the lifespan (Conway, Wang, Hanyu, & Haque, 2005; Conway & Haque, 1999; Larsen, Schrauf, Fromholt, & Rubin, 2002; Schrauf & Rubin, 2000, 2003). In both these areas one notes the same basic results in different cultures often with small but interpretable modifications for culture or major events in the cultural life of the society.

Here we view autobiographical memories as the products of component processes, with each process occurring in a separate behaviorally and neurally defined system. This is the basic-systems model of autobiographical memory summarized in Rubin (2005, 2006) and elaborated in a series of empirical studies (Greenberg & Rubin, 2003; Rubin, 1998; Rubin & Greenberg, 1998, 2003; Rubin, Schrauf, & Greenberg, 2003; Schrauf, 2000; Schrauf & Rubin, 1998, 2000; Rubin & Siegler, 2004). This model is distinctive eschewing the notion of abstract, amodal information in the brain and in tracing the components of autobiographical memories (visual imagery, auditory imagery, emotion, language, narrative, and others) to specific systems represented by identifiable areas in the brain. Thus, an autobiographical memory is the result of selectively re-integrating sensory, motor, emotional, and linguistic information from various brain centers into a narrative whole that corresponds to the original binding of some or all of that information at the encoding of the event (Rubin, 2006).

In previous work (Rubin, Schrauf, & Greenberg, 2003; Rubin & Siegler, 2004) we provided behavioral evidence that supports this basic-systems model by having participants retrieve memories and then complete rating scales representing the cognitive and metacognitive properties that make up the basic-systems. In particular, we examined how the component processes of autobiographical memory clearly represented in subsystems of the brain (e.g. visual imagery, auditory imagery, emotional reinstatement, language, narrative, etc) influenced two important metacognitive processes of autobiographical memory (recollection and belief). These latter, while not specific subsystems, are nevertheless critical dimensions of autobiographical memory.

Recollection is a metacognitive judgment about the extent to which a person re-experiences or relives an event upon remembering it. It differentiates memory from other states such as dreaming or imagining, but even within memory there is room for variability in recollection, as abundant research has drawn the distinction between ‘remember’ vs. ‘know’ judgments in episodic memory tasks (Gardner & Java, 1990; Tulving, 1985). Belief is a metacognitive judgment about the extent to which a person believes that a memory accurately represents an event from the personal past. Again, there is room for variability here as individuals commonly judge that memory is fallible and sometimes simply wrong. We found that higher recall of visual imagery, auditory imagery, and emotional reinstatement in memory predicted higher ratings of reliving the event (recollection), and that higher recall of the spatial setting of events predicted belief in the accord between what happened and the memory of what happened.

Because judgments of recollection and belief are metacognitive and not grounded in specific subsystems, we suspected that they might be more culturally variable than the component, cognitive processes represented in the basic-systems model. In this approach to culture and cognition, we assume that basic information processing systems, represented by identifiable areas in the brain, are panhuman but that the actual processing (integration and manipulation of information) is subject to cultural entrainment. This is analogous to other cross-cultural cognitive research such as that by Hedden, Park, Nisbett, Li-Jun, Jing, and Jiao (2002) that assumed that basic processing mechanisms such as processing speed and working memory would be present and measureable across cultural groups but that differences would emerge in processing outcomes as a result of interaction with culture. A similar assumption is made in social psychology, for instance, in the work of Markus and Kitayama (1991) who find cultural differences in cognitive processing related to the ‘self’ (but by implication find that the ‘self’ as a basic component of human being is pancultural). Thus, we sought to assess the cross-cultural variability of the metacognitive processes of recollection and belief in Japanese, Turkish, and American (U.S.A.) samples. Given that natural languages differentially draw attention to some aspects of experience and elide over others (Boroditsky, Schmidt, & Phillips, 2003; Slobin, 1996), we began by informally assessing whether the distinctions in the American ratings existed in Japanese and Turkish.

Consider the case of recollection. Just as the remember/know distinction makes sense in English, a similar distinction exists in Turkish between hatirlamak and bilmek and in Japanese between the verbs kiokushiteiru and shitteiru. Moreover, studies distinguishing remember from know judgments have been conducted in Japan and Turkey as well as the U.S.A. (e.g., Horiuchi, 2004; Tekcan, Goz, Yalcin, Akirmak, Serbest, Firat, & Rodoplu, 2002; Rubin, Hinton, & Wenzel, 1999). Thus, each of the languages spoken by participants in our study make similar distinctions between mnemonic information that is remembered vs. simply known. Nevertheless, it remains an empirical question whether and to what extent the component processes that predicted recollection in the American sample will predict recollection in the Japanese and Turkish samples.

On the other hand, there may be reason to predict that belief in one's memories— judgments that memories accurately represent originating events—is more variable cross-culturallly. In this regard it is of note that Turkish grammar has a special status for the narrative that impacts upon belief (Slobin & Aksu, 1982). There are two forms of past tense in Turkish, the –di form and the –miş (pronounced as mish) form. These two forms are verb suffixes for past tense that have different implications. The –di form is used for actions that are personally experienced or witnessed. The –miş form denotes hearsay and secondary sources as well as information which the individual inferred, or for which he or she was unprepared, or that has a quality of otherworldliness (e.g. myths and fairytales). These two grammatical forms distinguish the reliability of narratives. If a particular event is told in the –di form, the narrator has seen, experienced or felt it and therefore there is no reason to doubt the reality of the event. However, if the event is told in the –miş form, then the narrator has not seen it directly, suggesting that the reliability may not be ascertained.

We chose Japan, Turkey, and the U.S.A. for several reasons. One practical reason was the availability of research participants in these three areas who were sufficiently familiar with rating scales to take part in the study. Another reason for our choice was that we wanted a diversity of cultures. Anthropologists define ‘culture areas’ as regions comprising groups of societies with common environmental influences, geographical contiguity, shared history, and similar social structure (Burton, Moore, Whiting, & Romney, 1996). Clearly, these Japan, Turkey, and the U.S. represent historically different cultural and geopolitical regions. The three cultures also differ from each other on more quantitative measures (Hofstede 1980, 2001). Hofstede (2001), ranked 53 countries and regions on four cultural dimensions derived from factor analytic studies. His participants were marketing and service employees of a large multinational company. Most were middle-class college graduates. Additional data were collected from managers receiving post graduate training. Thus, the sample on which cultural diversity was measured was at least as educated as our own. Japan, Turkey and the U.S.A. had the following ranks on the four cultural dimensions: Power Distance (the degree of power difference between superior and subordinate as perceived by the less powerful): 33, 18.5, 38; Individualism (the level of distinction of the individual from the other in the social environment): 22.5, 28, 1; Masculinity (the importance attached to masculine goals as opposed to feminine goals) 1 32.5 15; and Uncertainty Avoidance (the need for clear, defined, and specified environment with a predictable future): 7, 16.5, 43. Uncertainty Avoidance is highly affected by the age of the respondents and once age is equated across samples, the spread in Uncertainty Avoidance becomes even larger with Japan, Turkey and U.S.A. being ranked 1, 22, and 37 in the 40 countries in which the age correction was done. Thus the ranks of the three cultures vary widely, implying that the three culture are, in fact, quite different.

McCrae et al. (2005) examined the similarities between 51 cultures on thirty Revised NEO Personality Inventory facet scores. A multidimensional scaling procedure where two dimensions were extracted and rotated to maximize correlations with Neuroticism and Extraversion revealed that the Japanese, Turkish, and American cultures are represented in separate clusters.

We are not claiming that differences described by global regions, Hofstede's four dimensions, or McCare et al.'s personality factors are the only differences in our three cultures, only that they indicate that we have sampled widely. Similarly, although we assume that differences in Individualism and Uncertainty Avoidance or personality dimensions like Extraversion are likely to provide us with considerable variability among our three samples on some of the properties we measure, we are not claiming that these differences are the only reason or cause for any variability that is observed.

In terms of autobiographical memory, a question like our scale of belief in the accuracy of a memory (i.e. I believe the event in my memory really occurred in the way I remember it and that I have not imagined or fabricated anything that did not occur, rated from 1 = 100% imaginary to 7 = 100% real) should be highly related to Uncertainty Avoidance. If one cannot usually trust one's own memories to be accurate, then the world would be a very uncertain place. Given the range of uncertainty avoidance across these three cultures, we should obtain a good range in our belief measures with Japanese participants being higher than Turkish participants who should be higher than the U.S.A. Similarly, the U.S.A. ranks highest in Individualism, whereas Japan is near the middle and Turkey ranks lower. This suggests that compared to Turkey and possibly Japan, participants in the U.S.A. should show less dependence on auditory imagery (at least to the extent that it reflects spoken language), language, and narrative compared to other component processes such as vision. That is, it seems logical to assume that participants from highly individualist cultures might rank their privately verifiable visual evidence over socially derived (e.g. “heard” evidence). Given these predictions, differences in mean values should follow, but these would say little about the mechanisms used to produce the belief in the accuracy of an autobiographical memory or auditory versus visual imagery. Thus, for questions of the relationships among variables we use correlations and multiple regressions.

Method

The basic procedure was set first in English and so we present the methods in detail for the U.S.A. sample, noting deviations from it for the Japanese and Turkish sample. All instructions and rating scales were translated and back translated by fluent bilinguals in order to ensure as close a translation of meaning as possible. The U.S.A. sample was analyzed in Experiment 1 of Rubin, Schrauf, and Greenberg (2003).

Participants

All participants were native speakers of the language group in which they were tested. For each culture we tested in excess of 50 individuals and then reduced the number to 50 to produce equal ns by eliminating those participants who had the most missing data either because they failed to report a memory to a cue word or failed to answer questions about a cue word. In the Japanese, Turkish, and U.S.A. sample there were initially 52, 57, and 55 individuals. The average age of the 50 Japanese participants was 19.30 (SD = .99, range 18 to 23); 26 were female. For Turkey, the average age of the 50 participants was 18.94 (SD = .71, range 17 to 21); 31 were female. For the U.S.A., the average age was 18.54 (SD = .76, range 17 to 21); 34 were female. Being university undergraduates effectively holds SES and education high so that any cross-cultural differences that do emerge should be due to culture and not these other factors.

Materials

Each subject was presented with a booklet consisting of a cover page with instructions, a double-sided sample page containing a sample cue word and 15 questions about the memory it cued, a dividing page, and then 30 more double-sided pages, each of which contained one of 30 cue words and the 15 questions referring to the memory it elicited. All questions, were seven point rating scales anchored at extreme values such as “not at all” and “completely”. Exceptions to the seven point scales were responses for the date of the memory and the judgment of whether the memory was of an event that occurred within a single day or was extended over longer than a day, or was the merging of several events The instructions began: “In this study we are trying to find out about the basic properties of autobiographical memories. To do this we will cue you with 30 words. For each word we will ask you to recall the first memory from your life that comes to mind and to think about it for a while before answering questions about the memory. There are no correct answers; we are just trying to document the kind of memories people have.”

Table 1 provides a list and brief description of our measures along with the brief names to which we refer throughout the text. The concepts tapped by these measures are the ones commonly considered and rated in the autobiographical memory literature (Conway & Pleydell-Pearce, 2000; Johnson, Foley, Suengas, & Raye, 1988; Rubin, Groth, & Goldsmith, 1984; Rubin & Schulkind, 1997; Sheen, Kemp, & Rubin, 2001). Each question is a proxy for one or more complex theories or theoretical ideas (Rubin, 2005). The full set of questions is given in the Appendix of Rubin, Schrauf and Greenberg (2003).

Table 1.

Variables Used in Study

Variable	Brief Description of Rating Scale Experiments
	Recollection and Belief
Reliving	I am reliving the original event.
Back in Time	I travel back to the time when it happened.
Remember/Know	I remember it rather than just knowing it happened.
Real/Imagine	I believe the event in my memory really occurred.
	Component Processes
See	I can see it in my mind.
Setting	I can recall the setting where it occurred.
Hear	I can hear it in my mind.
Talk	I or other people are talking.
In Words	It comes to me in words.
Story	It comes to me as a coherent story.
Emotions	I can feel now the emotions that I felt then.
	Reported Properties of Events or Memories
Importance	It is significant for my life.
Rehearsal	I have thought or talked about this event.
Once/Many	It occurred once at one particular time.
Merged/Extended	A merging of events versus an extended event.
Age of Memory	Please date the memory (month/day/year).

Note: All measures are seven point rating scales anchored at extremes, except for once/many. merged/extended, and age of memory, which is converted to days reported since the event.

Procedure

The participants were tested in large groups. They were read the printed instructions and asked to think of a memory to the word tree and then answer all 15 questions about it. Each of the 15 questions was discussed briefly and any questions from the participants were answered. The participants were then asked to recall a memory to each of the remaining 30 cue words in turn and, while they were thinking about that memory, to answer the 15 questions about it. This part of the task was self-paced. The complete procedure took between 50 and 90 minutes.

Results

For each group, the maximum number of possible memories was 1500 (50 participants time 30 memories each). There was little missing data. For Japan, Turkey, and the U.S.A., with the exception of the dating of the memories, the number of missing ratings for our various scales varied: between 7 and 9 for Japan with the exception of once/many which had 27 missing values; between 28 and 31 for Turkey; and between 2 and 6 for the U.S.A. All groups were asked to date their memories to the exact day, giving their best guess when possible. The Japanese students had 53 missing years, 222 missing months, and 858 missing days. We therefore restricted the Japanese dating to years. With this modification there were 53, 35, and 8 missing dates for Japan, Turkey, and the U.S.A., respectively. Whether the missing dates in the Japanese data is caused by a reluctance to guess and thereby provide inaccurate answers, a conjecture consistent with differences in Uncertainty Avoidance, remains an open question.

Means

As shown in Table 2, some rating scales, such as real/imagine, are generally high and others, such as in words, importance, and rehearse are generally low. In addition, there was considerable consistency in the ordering of ratings given to component processes. In all cultures, the highest ratings were given to either see or setting. These were followed by ratings of emotions. Next were the three ratings of auditory imagery and narrative (hear, talk, and story). The lowest ratings were given to in words. Yet there are still differences among the three cultures that are fairly large. For the seven point rating scales, the difference between the cultures with the highest and lowest rating is about 3/4ths of the average standard deviation within a culture. Thus, a person who scored at the 50^th percentile in the culture with the highest score would be at the 77^th percentile in the culture with the lowest score. For two scales, remember/know and setting, all three cultures differ by a Tukey's test; for two other scales, merge/extend and age of memory, none of the cultures differ. For the remaining 12 scales, there are 9 scales on which Japan differs from the U.S.A., 6 on which Japan differs from Turkey, and 3 on which Turkey differs from the U.S.A. More conservatively, of these 12 scales, there are 4 scales on which Japan is lower than the other two cultures (hear, talk, story, and once/many), 1 scale on which Turkey is lower than the other two cultures (rehearse), and 1 scale on which the U.S.A. is higher than the other two cultures (see).

Table 2.

Means and ANOVA for Differences in Memory Variables across Countries

	Country			F (2, 147)
	Japan	Turkey	U.S.A.
Variable	Mean (SD)	Mean (SD)	Mean (SD)
Recollection and Belief
Reliving	4.27 (0.97)a	4.59 (0.77)a,b	4.83 (0.80)b	5.52**
Back in Time	4.18 (1.07)a	4.54 (0.81)a,b	4.85 (1.09)b	5.70**
Remember/Know	4.37 (0.85)a	4.94 (0.72)b	5.69 (0.65)c	39.55***
Real/Imagine	6.19 (0.78)a	6.01 (0.73)a,b	5.75 (0.72)b	4.34*
Component Processes
See	4.83 (0.77)a	4.84 (0.77)a	5.37 (0.72)b	8.57***
Setting	4.80 (0.80)a	5.45 (0.80)b	5.85 (0.75)c	22.72***
Hear	3.50 (0.96)a	4.35 (0.80)b	4.23 (0.96)b	12.90***
Talk	3.55 (0.84)a	4.10 (0.88)b	4.34 (0.87)b	10.95***
In Words	3.21 (1.14)a	3.75 (0.79)b	3.43 (1.36)a,b	2.99
Story	3.52 (1.03)a	4.20 (0.83)b	4.49 (1.06)b	13.05***
Emotions	4.07 (0.85)a	4.41 (0.86)a,b	4.65 (0.91)b	5.47**
Reported Properties of Events or Memories
Importance	3.14 (0.84)a,b	2.98 (1.04)a	3.51 (0.88)b	4.20*
Rehearse	3.50 (1.04)a	2.99 (0.73)b	3.41 (0.81)a	4.83**
Once/Many	0.61 (0.16)a	0.71 (0.18)b	0.69 (0.15)b	5.60**
Merge/Extend	1.42 (0.26)a	1.54 (0.30)a	1.43 (0.27)a	2.53
Age of Memory1	1943 (743)a	1902 (899)a	1565 (779)a	3.28*

Means with different superscripts are different at p < .05 by a Tukey's test.

^*05

^**01

^***001

¹Age of Memory is the reported time between the event and recall in days

One interesting difference in the means that arises again in the correlational analyses that follow is that hearing appears more important in the Turkish sample. If one compares see to hear, two questions with very close wording that each ask about a different sensory modality, see is 1.33 and 1.14 higher in the Japanese and U.S.A. samples on a seven-point scale, respectively, but only .49 higher in the Turkish sample. A repeated measures ANOVA confirms the difference in mean level [F(1,147) = 313.75, p < .0001)], in culture [F(2,147) = 9.18, p < .001)], and their interaction [F(2,147) = 21.03, p < .0001)].

Multiple regressions predicting belief and recollection

In order to better understand why an autobiographical memory is recollected, why it is believed, and the relation between the two, we used multiple regression analyses to predict the reliving, back in time, remember/know, and real/imagine variables with 10 of the remaining 11 variables; merge/extended was excluded because it could not be calculated on the more than half of the memories that were rated as once on the once/many scale. The analyses were conducted between subjects using the mean ratings of the 50 subjects, and within subjects for each subject individually before aggregating over individual subjects’ analyses. Table 3 presents these results. For the between-subjects analysis, the equation with the highest r-squared that had only variables with beta weights significantly different from chance at the .05 level was chosen. For the within-subjects analyses, regressions were conducted for each subject separately and the parameters of the resulting equations were averaged (see Estes, 1956 for a discussion of averaging such linear equations; see Rubin, Schrauf, & Greenberg, 2003, 2004 and Rubin & Siegler, 2004 for other examples of using within an between subject analyses). A t-test was performed on the set of 50 beta weights associated with each independent variable to see if its mean was different from zero at the p <.0001 level. This procedure was first conducted with all 10 independent variables. Next, the independent variable with the smallest average beta weight was removed if it was not different from zero at the .0001 level. This step was repeated until all remaining beta weights were statistically significant. We used the .0001 level instead of the .05 level to arrive at a roughly equal number of predictors in the between- and within-subjects equations in order to facilitate comparisons between them. The change in p level from .05 between subjects to .0001 within subjects indicates the high degree of similarity among the regression equations of the individual participants.

Table 3.

Multiple Regressions

Dependent Variable	Type of Analysis		Beta Weights for Independent Variable											R2
			see	setting	hear	talk	in words	story	emotions	importance	rehearse	once-many	age of memory
Reliving	Japan	b/t	.49		.24				.31		−.22			.73
		w/i	.21	.17	.21				.29					.65
	Turkey	b/t	.44		.38		.21					−.14		.91
		w/i	.31		.33		.12		.19					.79
	U.S.A.	b/t	.50				.15		.45					.76
		w/i	.33		.27			.15	.17					.66
Back in Time	Japan	b/t	.58					.28					−.18	.58
		w/i		.32				.19	.37					.65
	Turkey	b/t	.51		−.40	.23		.63						.83
		w/i		.18			.24	.26	.20					.66
	U.S.A.	b/t	.39					.30		.24				.51
		w/i	.25		.13			.36		.14				.62
Remember/Know	Japan	b/t	.81										−.20	.70
		w/i		.40					.39	.16				.72
	Turkey	b/t		.51				.48						.79
		w/i	.14	.24			.30	.21						.67
	U.S.A.	b/t		.67				.35	−.27					.57
		w/i	.26	.31				.21					−.11	.63
Real/Imagine	Japan	b/t		.47										.22
		w/i	.33								.20			.35
	Turkey	b/t			.58									.33
		w/i		.34	.16			.29						.55
	U.S.A.	b/t		.49				.28	−.32					.30
		w/i	.19	.19				.26					−.20	.54

Note: b/t is for the between subject analysis, and w/i for the within subject analysis.

The between-subjects and within-subjects analyses ask different questions. The between analyses is a standard individual differences analysis. It asks whether the average ratings of people on our four dependent measures can be predicted by the average ratings of people on our ten independent measures. For example, we test for whether people who tend to relive their memories are also people whose memories have high visual imagery. The within analysis examines each individual separately, doing multiple regressions for that individual. These regression equations are then ‘averaged’. The question here is at the level of most theory in experimental cognitive psychology; that is, it asks what independent measures, within an individual, predict our four dependent measures within the same individual. For example, within each individual considered separately, are the memories that are relived most also the memories that have the highest visual imagery. Because the means used by the between analysis are subtracted out by the correlation formula of the within analysis, each form of analysis is based on information that the other does not use, and the two levels of analysis are independent of each other.

The regression equations, which are shown in Table 3, using the between and within units of analysis are fairly consistent across cultures. As in our earlier work (Rubin, Schrauf, & Greenberg, 2003; Rubin & Siegler, 2004), the reliving and back in time variables tend to have different predictor variables than the remember/know and real/imagine variables. First, although the see and setting variables are highly correlated, see more often enters into predictions of reliving and back in time; by contrast, setting more often enters into predictions of remember/know and real/imagine. The difference between being able to “see in my mind” and being able to “recall the setting where it occurred” could be a distinction between imagery and context, or between visual imagery and multimodal spatial imagery (Farah, Hammond, Levine, & Calvanio, 1988). Second, emotion or the correlated variable of importance enters more often into regression equations for the reliving and back-in-time variables and always with a positive weight, but enters into the two regression equations for the between-subjects analyses for the remember/know and real/imagine variables with a negative weight.

There is one major difference in the within- and between-subject multiple regressions, a difference that also occurred in our earlier work. The r-squared values are similar for most of the equations and if anything are larger for the between-subject regressions, but they are considerably lower for the between-subject regression of real/imagine. Real/imagine is not as well predicted as an individual differences variable as our other measures.

Comparing across cultures, there are few differences, but they are robust and either predicted or easy to interpret. Variables related to language and narrative (i.e., hear, talk, in words, and story) tend to have their largest contributions for the Turkish data, especially when compared to the Japanese data. Talk, in words, and story do not enter into any equations for the Japanese data except for back in time, where they enter for all cultures. In order to further examine these differences, the raw within subject correlations between the independent and dependent variables in Table 3 were examined, as opposed to the partial correlations used by the multiple regression equations. Correlations were calculated for each of the 150 participants individually. The Fisher's z transform of these correlations were submitted to one way ANOVA's with three levels of culture. These are shown in Table 4 along with the average correlation (calculated from the Fisher's z) for each culture. Of the 48 ANOVA's represented in the table, 25 are significant at the .05 level (whereas 2.5 would be expected by chance) and 13 are significant at the .001 level (which is the corrected table-wise .05 level; i.e., .001 = .05/48). Moreover the distribution of significant differences is far from random. Of the 16 analyses involving auditory imagery and language, 13 have significant F values. In all 16 of these analyses, the correlation from the Turkish sample is the highest numerically, and in 10 of these analyses the Turkish sample differs significantly from the other two samples. All four analyses involving emotion are significant. In contrast, of the 28 analyses from the remaining 7 independent measures, only 8 are significant and these show no clear pattern.

Table 4.

Mean Correlations among Independent and Dependent Measures of Table 3 Compared across Countries

	Phenomenological Report
	Reliving				Back in Time				Remember/Know				Real/Imagine
Variable	F (1, 147)	J	T	U	F (1, 147)	J	T	U	F (1, 147)	J	T	U	F (1, 138)	J	T	U
Component Processes: Visual-Spatial and Emotion
See	13.74***	66a	83b	70a	0.29	64a	64a	61a	0.20	67a	66a	65a	1.64	51a	59a	54a
Setting	2.31	66a	61a	58a	9.46***	70a	60a	53b	4.98**	75a	65b	65b	2.74	54a	62a	52a
Emotions	9.04***	60a	76b	63a	5.02**	70a	67a,b	58b	20.15***	76a	66b	52c	5.29**	42a	56b	47a,b
Component Processes: Sound and Language
Hear	18.08***	59a	93b	68a	2.63	54a	63a	57a	1.90	57a	64a	56a	10.57***	35a	56b	46a
Talk	6.86**	51a	64b	49a	1.47	53a	55a	48a	4.32*	54a,b	58a	45b	7.40***	33a	50b	38a
In Words	21.99***	40a	71b	38a	18.96***	41a	71b	37a	13.87***	50a	71b	37a	15.02***	30a	57b	32a
Story	21.67***	41a	69b	59c	5.64**	53a	69b	67b	6.83**	50a	68b	58a,b	20.72***	30a	62b	58a
Reported Properties of Events or Memories
Importance	0.33	48a	48a	45a	1.30	52a	47a	44a	10.21***	56a	43b	33b	0.98	35a	41a	36a
Rehearse	1.76	41a	48a	41a	0.85	46a	42a	40a	5.48**	50a	43a,b	34b	2.06	40a	40a	31a
Once/many	4.46*	00a,b	03b	11a	2.65	00a	00a	10a	3.05	01a	00a	11a	1.43	04a	06a	12a
Merge/Extend	0.67	09a	07a	19a	1.01	20a	06a	18a	0.29	18a	10a	14a	1.42	27a	09a	14a
Age of Memory	5.65**	–33a	49b	43a,b	0.54	–37a	42a	41a	2.24	–33a	43a	40a	4.57*	–31a	42a	47b

J is Japan, T is Turkey, U is U.S.A.

^*05

^**01

^***001

for merge/extend the df in the denominator are 116, 115, 116, and 102. Means with different superscripts are different at p < .05 by a Tukey's test.

Discussion

There are three key findings. First, there is a marked similarity across cultures in the relative values of the means and correlations of the 15 scales and in the multiple regressions. These include: real/imagine was predicted better at the within than between subjects level of analysis, remember/know ratings were predicted in multiple regression equations in ways that were more similar to measures of real/imagine than recollection, and setting was a better predictor of real/imagine whereas see was a better predictor of recollection. These results also occur in other samples from the U.S.A. Second, within that pattern of similarity, as expected, the Japanese sample expresses less uncertainty in the accuracy of their memories. Their high rating of real/imagine is in contrast to their generally lower ratings on other measures. The order of real/imagine across the three cultures follows the order of Uncertainty Avoidance as expected: Japan, Turkey, U.S.A. Given the many differences among these cultures, we cannot infer that Uncertainty Avoidance caused the differences in real/imagine, but the order of real/imagine is consistent with a prediction based on Uncertainty Avoidance. Third, the Turkish sample has higher correlations of measures of auditory imagery and language with measures of recollection and belief and higher mean rating of auditory imagery relative to visual imagery. This is consistent with the emphasis shown in Turkish grammar on whether one learned about an event through language or direct observation. Thus it appears that the same basic processes and relations are operating in all three cultures, but that there are small but measurable and understandable differences in means and in the basic relations among the variables as measured by correlations.

The clearest of these correlational differences among the cultures is the greater emphasis on sound and language in the Turkish sample vis-à-vis the other two cultures. In the introduction, we suggested that members of individualist cultures could be expected to rate the visual intensity of their memories higher than members of collectivist cultures because the former might rely more on visual evidence, while members of a collectivist culture would rate auditory imagery and narrative character of their memories due to the salience of social links. Thus, participants from Turkey, which is the most collectivist culture of the three (Hofstede, 2001), show higher rankings on language and narrative variables. However, one might have expected the U.S.A. to be the outlier with Turkey and Japan being closer together. One possible explanation for this may be found in work on narrative development by Minami and McCabe (1995). This work shows that the narratives of Japanese children are composed of brief and plain episodes of past events, and that Japanese mothers interact with their children in ways that encourage their production of these shorter narratives. This structure is not solely a discourse structure; instead it reveals a cultural norm for a socially situated activity (Labov, 1972). This Japanese enculturation into a more terse narrative style may counteract the more elaborate social and linguistic effects found in the Turkish sample (Kuntay & Ahtam, 2004). At an early age, the narratives children in these cultures showed differences such as use of evaluative remarks. Turkish children often tended to use evaluative remarks which did not emerge in the narratives of the Japanese children (Kuntay & Nakamura, 2004). Thus it may be that elaborated conversation plays a greater role in Turkey and thus in autobiographical memories collected from Turkey.

More speculatively, research indicating cultural differences in personality traits coincide with the differences in emphasis on sound and language. Research comparing cultures on the cultural averages of the five personality factors measured by Revised NEO Personality Inventory shows about one standard deviation of difference between Extraversion mean for Japanese in comparison to Turks and Americans (McCrae, 2002). The higher ratings of auditory information by the Turkish and American samples may be related to the elevated auditory component in the interactions of people who are highly extraverted. Furthermore, cultural and personality differences in the contents of memories and the need for coherence with self-images may impact upon the type of imagery associated with the memory of the event (Conway, 2005).

The basic-systems model of autobiographical memory reads the phenomenology of remembering (e.g. “see it in my mind,” “feel now what I felt then,” “reliving the event,” etc.) onto the neuropsychology of remembering (e.g. “visual cortex,” “auditory cortex,” “frontal lobes”, etc) in a way that acknowledges that the biological architecture of memory and the fundamental sensory accompaniments are pancultural universals. Nevertheless, our research suggests that different components in the system may be activated differentially in stable patterns that differ cross-culturally. Specifically, the question we addressed in our analysis was: how do ratings of sensory and other features and judgments of metacognitive processes interact in predicting Japanese, Turkish, and American undergraduate judgments about two other key properties of memory: the sense of reliving one's memories and one's belief in the veracity of one's memories. Results indicate that the Turkish pattern differs from the Japanese and American patterns in that language and narrative variables are more implicated in the Turkish students’ sense of reliving and believing their memories than they are in Japanese and American's sense of reliving and belief in their memories. Thus, while the underlying mechanics of memory may be cross-culturally identical, the processing taking place seems to differ. It is not simply a difference in how Turkish individuals assess their memories; the reliving, back in time, remember/know and real/imagine judgments are post-hoc assessments of a memory already retrieved. It is also a difference in how these individuals experience their memories. That is, the sense of reliving and experience of being “back in time” are constitutive of on-line vivacity. While such experiential reliving of memories is common across the three cultures (and arguably universal), its mental provenance is not. For Turkish students, it is grounded more on language and narrative processes than for Japanese and American students. Why this should be so is certainly interesting, but that it is so suggests that we need to continue refining our understanding of how cognition and the sociocultural environment interact in cognitive processing. Put simply, it is not that people from different cultural backgrounds think different thoughts, rather they think differently. Our research suggests one way of understanding how this might work at a processing level.