Although parents are of primary importance when children are younger, there is an important psychosocial change that takes place through childhood and especially adolescence: friends become increasingly important as children get older [9, 10]. This change in the relative importance of parents and friends is reflected by changes in the number of memories that individuals can readily access about parents versus friends across childhood and adolescence. Conway and his colleagues [11, 12] argue that the memories that are most meaningful both at the time of event occurrence as well as at the time of retrieval are those that are elicited by a time-limited task. When such a task, termed a memory fluency task [13, 14], was used with university-aged young Italian adults for independently eliciting memory reports about parents or friends across various periods of life (preschool years, elementary school, middle school, and high school and University years), it was found that memory reports about parents were comparatively more numerous in the preschool years and the frequency of parental memory reports remained constant across increasing child age [14]. However, the frequency of memory reports about friends increased linearly and exceeded those of parents in the later periods of life, thus reflecting changes in the importance of psychosocial relationships with parents and friends as children grow up. Reflecting a similar pattern with a different type of data, Larson and his colleagues [15, 16] time-sampled children and adolescents to investigate how much time they spent with various companions. They found a dramatic decrease in amount of time spent with parents as the age of the youth increased, as well as more time spent with friends.

The nature of parent-child and friendship relationships also changes over time. Parent-child relationships are inherently hierarchical and complementary. They become more characterized by conflict and negative affect as children get older, particularly as children approach adolescence [10], although affect during parent-child interaction has been shown to become more positive again in late adolescence [16]. The increase in negative affect is partly because adolescents are struggling with establishing personal identity and social autonomy [17]. Adolescents report more conflict as well as more intense anger in interactions with parents [18] than with friends [19], and display more positive affect when they are with their friends than with their parents [16]. This asymmetry is related to relationships with parents being less egalitarian as well as involuntary. Moreover, serious conflict does not terminate the relationship. In contrast, conflict with friends can threaten the existence of a relationship, and consequently, friendship relationships are more symmetrical. And as individuals get older, friends become increasingly important sources of emotional support and self-esteem [20, 21]. These changes in the affect of parent-child versus friendship relationships are reflected in young adults’ affective ratings of their memory reports about parents and friends across age. Peterson et al. [14] found that the proportion of memories about parents that were rated as negative increased over age as individuals recalled memories from their preschool through adolescent years, whereas there was no change in the proportion of negative memory reports when recalling memories about friends.

The relationships that adolescents and young adults have with their parents are often moderated by gender. A variety of research studies suggest that, in general, females have closer and warmer relationships with their parents in their adolescent years than do males [22]. They are more involved within family life [23], talk more about their plans and feelings with parents [24], and have more effective communication strategies with parents than do males [25]. They also seem to feel stronger obligations to support and respect their parents [26]. In contrast, parent-son relationships at this age seem to be more frequently affectively negative in tone than are parent-daughter relationships [27]. Males often use more confrontational tactics during conflict with parents [18] and in turn, parents are more likely to use coercive strategies with sons [27]. These differences in parent-daughter versus parent-son relationships are also reflected in the memory reports of young adults who are engaged in a memory fluency task. Peterson et al. [14] found that although memory reports that involved parents became increasingly likely to be rated as negative as young adults reported memories from older periods of their lives (preschool through late adolescence), this was particularly true for males. Fully half (51%) of the readily-accessible memories males reported about parents in a memory fluency task were negative when they recalled their high school and university years, versus only 37% of those by females.

Likewise, gender can play a moderating role in relationships between friends. Females typically have friendships that are more likely to emphasize intimacy, affection and self-disclosure [28, 29]. Their friendships are more likely to involve intense dyadic relationships with a small number of friends, i.e., they are more exclusive [30]. They spend more time in discussion, particularly in sharing their feelings [31, 32]. In contrast, males are more likely to focus on activities and doing things together in their interactions with friends, and are less likely to talk about feelings and their private lives. Their friend interactions are more likely to involve groups as well as competition [33]. In terms of memory reports in a memory fluency task, Peterson et al. [14] found no differences in affect for memories of friends between males and females. However, males’ memory reports about friends were more likely to be episodic, unique events rather than repeated events, which was interpreted as reflecting males’ greater orientation toward doing activities with friends rather than engaging more in interaction focused on talking and sharing feelings.

Interactions between interpersonal relationships and memory reports have been highlighted above, but the measures of memory that were used in much of this research typically took a broad-brush approach. For example, some described the sheer number of memories about different relationship partners that young adults could readily retrieve in a memory fluency task, or the overall affective tone of those memories [14, 34]. However, memory narratives are more complex than this.

One linguistic property that has been studied in relation to qualitative aspects of interpersonal relationships is pronoun use, and in particular, first person singular and plural pronouns (‘I’ and ‘We’). Proportionately greater use of ‘I’ has been found to be associated with poorer relationship quality within couples in some research, such as lower marital satisfaction [35] and low involvement in dyadic communicative interaction [36]. These researchers, along with Williams-Baucom, Atkins, Sevier, Eldridge, and Christesen [37], concluded that proportionately greater use of ‘We’ and lesser use of ‘I’ seem to be linguistic indicators of higher quality and stability of a couple’s relationship. Differential use of ‘I’ and ‘We’ has also been found to be associated with quality of friendship relationships [8]. When young women were asked to write a detailed narrative about a memory that involved their best friend, those with more qualitatively positive friendship relationships used proportionately fewer first person singular (‘I’) and more first person plural (‘We’) pronouns. In fact, the higher the degree of rated social support women experienced in their friendship relationships, the fewer ‘I’ pronouns they used, and conversely, the greater the degree of negative interaction they identified in their friendship relationships, the more the pronoun ‘I’ was used. The findings for young men were quite different, however: men who rated their friendships as more qualitatively positive had a proportionately greater use of ‘I,’ which the authors interpreted as consistent with males’ friendship relationships being less characterized by intimacy than females’ and more likely to have dominance hierarchy overtones [33, 38-40].

Prior research has also assessed emotion in personal narratives, by either rating the overall emotional tone or by individually counting and categorizing the specific emotion words that are used. These emotion words describe the emotions experienced by the narrator as well as others during the narrated event. They have been shown to reflect the overall emotional tone of experiences that are explicitly positive or negative [41], as well as are effective indicators of emotional tone in parent-child interactions [42] and romantic relationships [43].

In terms of young adults’ memory reports about parents at various earlier points in their lives, as we stressed above, memory reports of experiences that involve parents become more negative as the rememberer increases in age from preschool-aged through adolescence [14]. Furthermore, those individuals that have parents who are more supportive and involved with their children have relatively more positive memory reports about parents than do those individuals with more conflictual and less positive parent-child relationships [7, 34, 44]. However, these researchers assessed the overall emotional tone of short memory reports that had been gathered during a memory fluency task rather than assessing the number and emotional valence of specific emotion words in individuals’ memory narratives about parents – which are considerably longer than the short reports elicited by a memory fluency task. This is a more differentiated look at affect because a particular memory narrative could contain multiple emotion labels as well as a mixture of different types of emotions. Tani et al. [8] have done such a detailed analysis of specific emotion words in young adults’ memory narratives about their best friends. Using the Language Inquiry and Word Count procedure (LIWC) developed by Pennebaker, Booth, and Francis [45] which provides a lexical analysis of text, Tani et al. [8] found that both males and females used more negative emotion words in their memory narratives about their best friends when their friendship relationships were characterized by an unequal power balance in the friendship. However, to our knowledge no one has applied such a lexical analysis of emotion words to memory narratives of parents, nor compared narratives of parents versus friends.

To sum up, because language, as Chung and Pannebaker [46] observed, is the currency of most human social processes and we use words to convey our emotions and thoughts as well as to tell stories and to understand the world, it is important to study how language is used to communicate memories. This is the purpose of the current research. Furthermore, Pannebaker's studies demonstrated that not only content words can be indicators of mental states and social relations, but function words such as pronouns can as well.

The purpose of the present study was to compare the narratives young adults provide about memories that involve their parents to those involving their best friends for events that occurred during their late adolescence or early adulthood (high school and university years). Since there are substantial differences in the number of readily accessible memories about parents versus friends as well as the overall affective valence of these two types of memories when short reports are elicited by a memory fluency task [14], we anticipated differences when a more fine-grained analysis of the language of longer memory narratives was conducted. To avoid interference effects between the two types of memory narratives, participants were asked to provide only one memory narrative, and half of the participants were asked to do so about their parents and the remainder about friends. The memory narratives were analyzed using the LIWC procedure [45] for a number of types of words. In addition, the memories were rated in terms of their overall affective valence.

Hypotheses. The LIWC provides information on the ratio of emotion words to all words, and its listing of the specific words used in the narratives allows one to assess the affective valence of these words. We expected there to be differences in the ratios of emotion words to all words between parent and friend narratives, and that the valence of these emotion words would differ. We predicted that narratives about parents would have more negative emotion words because Tani et al. [8] found that the number and emotional valence of emotion words are related to relationship quality, at least for friendships. Although they did not assess narratives about parents, it makes sense that emotion words in narratives about parents versus friends would differ since short memory reports (elicited by a memory fluency task) about parents are more negative than those about friends from this period of an individual’s life [14]. Furthermore, relationships with parents are more likely to be conflictual and affectively negative at this age [10, 17] and include more intense anger [18]. We expected that the proportion of negative emotion words would be especially high in males’ narratives about parents, in keeping with the gender differences in parent-child relationships discussed above. In addition, when the overall affective valence of narratives is rated, it is predicted that narratives about parents would be more negative than those of friends, especially for males.

A number of other types of words were also assessed because they reflect important components of narratives [1, 2]: number of words, first person singular and plural personal pronouns and third person singular and plural personal pronouns, mental states such as cognitions, as well as causal, temporal and adversative connection words. In keeping with the research on use of pronouns described above, we hypothesize a greater use of ‘I’ in narratives about parents and greater use of ‘We’ in narratives about friends, as well as longer narratives by females. In terms of words denoting cognitions and connectives, given the fact that to our knowledge, no prior research has explored how these types of words differ between narratives about friends and parents, we have no a priori hypotheses about how the use of them would be related to the two types of narratives.

A total of 177 university students from the University of a city in the centre of Italy (89 males and 88 females) were recruited. Students were 18 to 24 years old (M = 21.5 years; SD = 1.8); males’ mean age was 21.7 years (SD = 2.1), and females’ mean age was 21.3 years (SD = 1.5). The majority of these participants came from the centre of Italy (82.7%), with the remainder from Southern (9.2%) or Northern (8.1%) Italy. All participants came from families of middle or high socioeconomic level with more than 65% of their fathers and 63% of their mothers having a high school diploma or university degree. As well, 61.5% of the participants had at least one sibling and 72.1% currently lived at home with their parents. The students were divided randomly into two groups: the first consisted of 94 participants (42 females and 52 males), the second consisted of 83 participants (46 females and 37 males). The two groups were also similar in age (mean = 21.7, SD = 1.8 vs. mean = 21.5, SD = 1.8), geographical provenance (80.6% vs. 83.4% came from the centre of Italy), family education (first group's father's and mother's high school diploma were respectively 60% and 58%; second group's ones were 67% and 63%), and family structure (in the first group 60% had at least one sibling and 70% lived at home, in the second the percentages were 62% and 73% respectively).

Participants were recruited in class during courses at the University. They were informed of the project’s aims and were told that the experimental procedure required them to write about their past experiences and that it would last not more than 30 minutes. Participation in the research was voluntary and involved about 95% the students enrolled in the courses. All participants filled out a demographic form requesting information about their gender, age, place of birth, parents’ educational level, if they had siblings, and if they lived with their parents. No incentives for participation were given. Data were collected in class collectively by two researchers trained in narrative studies. The research was conducted in accordance with the guidelines for the ethical treatment of human participants of the American Psychological Association. Prior permission was obtained from the University Dean and President as well as each course professor. Participants provided their individual consent and could withdraw at any time.

A memory narrative task was given that consisted of two parts. In the first part a memory fluency task was used. Participants were given a sheet of paper with separate lines labeled for memory 1, memory 2, memory 3 and so on. They were then instructed: “Recall as many memories as you can that include your best friends (Group 1) or your parents (Group 2) from Middle School onward. You have three minutes to write these memories on this sheet of paper.” In the case of the first group, it was clarified in the instructions that they were not to choose as friends either those individuals with whom they had been, or were, romantically involved, or family members. This procedure is recommended by Becker, Johnson, Craig, Gilchrist, Haigh, and Lane [47]. However, we did not restrict participants’ choice to same-sex best friends. The time-limited recall task (timed by the researcher) was used because prior research has argued that the memories that are more readily accessible are those that are meaningful not only at the time of retrieval but also at the time the events occurred [11, 12]. Moreover, as Conway and his colleagues have further argued, the most accessible memories from a particular period of one’s life are best measured by providing subjects with a limited amount of retrieval time. For more information about the Memory Fluency Task see Peterson et al. [14].

In the second part of the task a sheet of paper was given to the participants and they were told: “Please select the most significant memory from those previously recalled regarding your relationship with your ‘best friends’ (first group) or ‘your parents’ (second group), and describe it in writing in detail.” The two sides of this sheet of paper had a maximum of 40 lines. This detailed memory task allows one to assess the narrative properties of a significant memory report. Participants were given 15 minutes to do this narrative task (timed by the researcher).

All narratives about friends and parents were first transcribed in txt format. In order to describe the main features of their content, all narratives were analyzed according to the content categories described below.

Two raters who were blind to the purposes of the study independently read all narratives and, using the systems described above, classified them in terms of each of the qualitative features described above. Disagreements were resolved through discussion. Cohen's K coefficients for agreement were respectively: Number of characters: K = .91, Theme: K = .92, Age at the time: K = .81, Episodic vs. repeated event: K = .95, Reportability: K = .87.

Narratives were analyzed using the Language Inquiry and Word Count procedure (LIWC) [45] for a lexical analysis of the text in order to quantify linguistic dimensions of these narratives. The LIWC program electronically processes text files one word at a time, matching the base form of words to an extensive dictionary of over 2290 word stems, and provides the percentage of words in several linguistic, emotional, and cognitive categories, regardless of the content of the events that are reported. A frequency count of the total instances of target words from each category is provided, and this count is then divided by the total number of words in the text to control for individual differences in verbosity. Thus, scores reflect a percentage of word matches in each category. The LIWC dictionary has been demonstrated to be reliable and exhaustive in its counts, categorizing approximately 85% of specific words used in a wide corpus of narratives and has been utilized by several narrative researchers [45, 52, 53]. In the present study, we used an Italian version of this dictionary that was elaborated and used on an Italian sample by Smorti, Pananti and Rizzo [54]. Given the hypotheses of this study the following categories included in the LIWC dictionary were examined:

Data analyzed from language categories 2 - 4 above were percentages of the target words in proportion to the overall number of words calculated by LIWC software.

The memory narratives produced by the two groups of participants were compared on the qualitative characteristics described above to see if there were differences between the two types of narratives. Likewise, potential gender differences were assessed.

Episodic vs. repeated events Narratives were mostly episodic in both groups (86% and 84%, chi square = .118, p = ns). The two genders were quite similar (90% and 85% for females and males, respectively, chi square = .99, ns]