What is a Friend ? An Exploratory Typology of the Meanings of Friendship

In this article I explore the contemporary normative meanings of friendship, unpacking the subject through two different questions: “what is a good friend?” and “what is an intimate friend?” Drawing on survey data from a national representative sample (n = 1142), the topic is explored in the context of a southern European country (Portugal) that represents an interesting case study, for its characteristics of late, though abrupt, entrance into late modernity. Statistical analysis of the results enabled the construction of an exploratory typology of representations of friendship, according to the meanings ascribed to friends: family-oriented; trust-oriented; self-oriented; and presence-oriented. Results inspire a two-folded interpretation. On the one hand, they point to a pervasiveness of hegemonic representations such as friendship as trust and self-disclosure, namely among the younger and more educated. On the other hand, they highlight the pervasiveness of kinship ties in the definition of friendship, namely among the elderly and less educated. This suggests that patterns of suffusion may not only refer to more individualized and plural arrangements of personal life, but also to the persistence of more traditional representations and practices, characterized by an ideological commitment to the family in its more institutional forms.


Introduction
This paper focuses on a very simple question: what do people mean when they talk of friendship?This question is unpacked in two different ones: (1) what do we mean by a good friend?(2) what do we mean by an intimate friend?Both questions are put in the context of a southern European country-Portugal-with a particular historical and cultural pathway regarding family and personal life, in which research about this topic, with a specific focus on friendship, has been practically inexistent.
Despite remaining underdeveloped, social sciences have approached the topic of friendship for several decades.However, attention paid to the subject has been irregular and dispersed through many different foci of analysis, with a wide variety of methodological approaches and theoretical backgrounds, which makes it very difficult to provide a comprehensive account of the social norms and rules that shape contemporary representations of friendship.An additional problem is that, as some have noticed [1], the very concept of "friend" is often taken for granted, relegated to an "essence" plan as if it required no definition at all.

The Meanings of Friendship: A Brief State of the Art
An important contribution to overcome this gap has been made by psychologists and social psychologists, namely within the scope of the study of interpersonal relations.For instance, Fischer [1] analysed data from a cross-sectional study of 1050 adults living in northern California in 1977.His respondents revealed an impressive set of relations (19,417), which Fischer unpacks into friends (the most common category, suggesting an undifferentiated use by north-Americans), relatives and other associates."Friend" seemed to be a residual label, applied to everyone to whom no more specific title was available, oriented to people of the same age and, in terms of contents of relations, oriented to sociability, to the "ones in which people visited, went out together, discussed shared past times, participated in an organization together" ( [1], p. 306).Maybe one of the most interesting contributions of this work is the possibility of a distinction between friends and close relationships, the latter more associated with intimacy, involving discussing personal matters, asking for advice or material exchanges.
Parks and Floyd [2] focus precisely on this concept of closeness.They compare the meanings given to closeness and intimacy within friendship and conclude that the former appeared to be a richer and more inclusive term than the latter, with respondents generating more meanings for it, as well as reporting a greater variety of relationships as close.Intimacy emerges therefore as an important topic to understand friendship in these studies.In 1992, Monsour [3] had already explored it, unpacking its meanings in cross-and same-sex friendships.He concluded that intimacy is multidimensional, and comes up with an array of seven most frequently mentioned meanings, with different distributions according to the type of friendship (cross-or same-sex, male or female): self-disclosure, emotional expressiveness, unconditional support, physical contact, trust, (sharing) activities and sexual contact.From a sociological perspective, Jamieson [4,5] has made an important contribution in defining intimacy as not strictly related to physical contact, particularly sexual contact, but rather "to the quality of close connection between people and the process of building this quality"."Closeness" can be emotional, cognitive, but also physical and involving bodily intimacy.The point is that an "intimate relationship need not be sexual and both bodily and sexual contact can occur without intimacy" ( [5], p. 1).
In many studies, the dyadic relation is elected as the main analytical focus.For instance, Johnson et al. [6] and Becker et al. [7] make a review of several studies about this subject, most of them taking the dyadic friendship as a reference and analytical tool.In these quantitative research designs, friendship is conceptualized in terms of level regarding closeness, self-disclosure, interdependence, instrumental and social support, shared interests, sharing of affection, among others.In all cases, friendships are never task-oriented, but rather person-oriented: their goal is to pursue intimacy and enjoyment, hopefully for both parts.Many of these studies adopt a three level typology of friendship, which is also adopted by Becker et al. [7]: casual, close and best friends.These levels are strongly and positively correlated with commitment levels.
Contrary to such a "dyadic approach", sociologists have tried to put friendships in context [8].By context Adams and Allan [8] mean the conditions external to the development, maintenance and dissolution of friendships, all the elements that surround them, being extrinsic to them, and not inherent.This perspective is central to move forward the dyad-centred one, towards a representation of friendships as the result of wider and more complex networks and levels of social relations.The importance of commitment in the definition of friendship has also been underlined by the work of other sociologists.According to Spencer and Pahl [9], to answer the question "what is a friend?"two important criteria become crucial: the nature of the bond (chosen vs. given) and the degree of commitment.Although the authors sometimes uncritically assume that "friends are essentially chosen relationships" ( [10], p. 82), they argue that a dichotomous and sharp contrast between given and chosen ties does not account for the diversity and complexity of real friendship practices.This way they explore the complex links and inter-relations of friends and family, which they call suffusion.These are the cases in which friends are seen as family, and vice versa, accounting for a blurring of the boundaries between supposedly given and chosen ties.Family members are not always seen and treated as "given", but may be seen as "chosen" (e.g., brothers in whom we confide).Likewise, friends may not always be perceived as "chosen", but may be ascribed "given" qualities (e.g., friend from early childhood).
Probably because the ways in which people experience their own friendships is so much embedded in hegemonic discourses, sometimes the doubt remains whether they are referring to attributes of actual friends/relations in their narratives of practices or to normative qualities of friendship.As Smart et al. [11] put it: a "strong moral regime of friendship" appears in discourses about (difficult) friendships, particularly those of women.Also, when describing the nature of commitment within friendship, Spencer and Pahl [9] underline that "when people talk about what it means to call someone a friend, it seems that part of their conception (...) still implies a certain kind of morality."In the discourses of their respondents, qualities like loyalty, constancy, caring, acceptance, honesty, truthfulness or reciprocity emerge as rules that draw the (moral) boundaries of friendships.
The persistence of these moral rules is of utmost importance since they translate hegemonic abstract ideas about what is relevant, useful or desirable in a given culture or society, and insofar as they serve as guidance regarding the socially acceptable behaviour, in a given context.While no direct relationship can be assumed between normative meanings of friendship and the ways people deal with their actual friends, the fact is that friendship expectations become a standard to rate both current and new friendships, thus influencing factors such as the evaluation of friendship quality or satisfaction [12].According to Hall [13], the qualities and behaviours of an ideal friend have been described by at least 37 studies; his meta-analysis of these studies identified four main dimensions of friendship expectations: symmetric reciprocity, agency, solidarity and communion.More recently, Hall [14] subjected the existing measures about the topic to confirmatory factor analysis in order to identify the latent structure of friendship standards.He reached a six-factor model, in which friendship ideal standards may be summarized as symmetric reciprocity, agency, enjoyment, communion, similarity and instrumental aid.
Also Allan [15] describes the core elements of contemporary normative definitions of friendship.Firstly, friendships are pictured as egalitarian, non-hierarchical and reciprocal.Friends are not supposed to have very different power or authority towards each other, even if they occupy very different positions within the wider social and economic structure.However, because this is difficult, "birds of a feather tend to flock together", and friendships tend to develop between people who have common social backgrounds, interests and lives, in a process of homophily [16], similar to the one we find regarding romantic relationships and marriage [17,18].When friends distance themselves in the social and economic structure, friendships often dissipate.Equality also means a balance in the emotional, symbolic and material exchanges, usually implicit rather than explicit.Another key feature of contemporary friendships is that they expected to be personal, i.e., not based on social roles or positions within the social structures.Friends are supposed to relate to each other in an authentic way, exclusively on a free and voluntary basis and taking as sole reference the quality of the relationship.Finally, friendships are expected to be non-instrumental, in the sense that they are not viewed as a means to achieve an end.Although this does not exclude social support (friends are supposed to help each other, that's part of the major modern narrative of friendship), utility is not on the basis of the ties between friends.Exchanges of resources are a result of friendship, not their ultimate end.All these features make friendships very demanding relationships, difficult to manage.
Such a prolific research has shown that norms and ideals of friendship constitute a relevant topic in and of themselves: they represent standards through which this kind of personal relationship is set as "normal" in a certain culture and context, against which current and new friendships are evaluated.However, the majority of existing quantitative research on the topic is based on non-probabilistic samples, mostly of university students from northern American origins, or of individuals from other country origins, but collected online.In both cases, the question of applicability of results to other social groups and cultural contexts remains unsolved.

The Relevance of the Portuguese Social-Cultural Context as a Case Study
In this article I aim precisely at exploring the prevalence of such moral qualities and norms of friendship in the context of a southern European country, Portugal, using a nationally representative sample.The relevance of this particular empirical context is linked to this country's specific path, characterized by a late, though quite abrupt, entrance into late modernity, with very rapid changes regarding certain social indicators (such as rise in education levels), as well as values (including those related to gender roles, family and personal life).It is also a country much characterised by contradictory practices and values regarding personal and private life, in particular: on the one hand, the persistence of "an explicit ideological commitment to the family", with strong intergenerational obligations, though unequally distributed and based on values of residential autonomy [19]; and on the other hand, a dual earner model within couples, shaped by greater gender equality, as well as rapid changes towards family diversity in terms of practices (divorce, post-divorce families, same-sex families).Processes of individualization have likewise shown growing expression (e.g., rising number of individuals, namely of working-age women living alone; higher educational, professional and residential mobility among the younger generations) [20,21], alongside the pluralisation of personal and intimate life (e.g., diversity of lifestyles, diversification of the number of sexual and conjugal partners across the life-course, diversity of family configurations) [22].
All these tensions make Portugal an interesting case study for exploring the changes that are usually at the frontline of modernity-including those occurring within the sphere of intimacy and personal life.What can we learn from the specific reality of such a context that may add knowledge to the set of studies conducted in other societal contexts?What differences, and what similarities, does this subject present under different cultural backgrounds?Can we speak of specific cultural expectations and representations of friendship?My hypothesis is that the particular position of Portugal in the context of western and European societies plays a role in the plural and diverse ways through which personal relationships and friendships are built in late modernity, challenging assumptions of a universal linear path towards more individualized patterns of living.Sociologists have described Portuguese society as one that proceeds not at a single speed, but at several and simultaneous speeds [20].A country that underwent enormous structural changes since the 60's, ranging from demographic composition and spatial and urban configuration to educational and professional structures, political democratization, opening to new lifestyles and wide diffusion of modern values [20].A country which has in the last 50 years become closer to other European countries in indicators such as fertility rate, ageing, female labour market participation rates or the rise of urban middle classes.However, these trends of modernity coexist with the persistence of traditional traits, which have been said to result precisely from a deficit of modernization [20].Among these, one of the most important ones, due to its impact in socio-economic development, is the low level of education and professional qualifications, which, though having improved drastically for the last 40 years, is still below the European and Organization for Economic and Co-Operation Development (OECD) average [23].Such structural gaps persist still today, when Portugal appears unexpectedly closer to Eastern European countries in certain aspects, such as social values (e.g., mutual trust).Almeida [21] talks cautiously about a "theory of abrupt transitions", according to which the very rapid implementation of societal changes, after a drastic shift to democracy, produces specific effects, namely related to the positive (or negative) expectations of citizens.These are some of the specificities that make Portugal an interesting locus of analysis of the on-going transformations of personal and intimate life, namely friendship, in late-modernity societies.To put friendships in the context of "abrupt transition" societies (Portuguese or other) may bring new perspectives about the general impact of late-modernity transformations in personal and intimate life.
My main objectives may therefore be summarized as: (a) to identify the normative meanings given to friendship; (b) to build a typology of the meanings of friendship; (c) to identify groups of respondents according to the meanings they give to friendship; (d) to characterize them in terms of social composition and their prevalence in the Portuguese society.This is therefore a descriptive and exploratory study that aims to explore how the meanings given to friendship are structured and distributed across the social fabric.
Cultural norms and ideals vary widely not only across cultures, but also within the same culture, according to factors such as social class, gender, age, education or stage of the life-course.Many studies have shown how normative discourses about friendship also vary according to different factors and contexts, such as the sex of the respondent and the friend (e.g., [2,3,6,24]; for a review, see [13]), the stage of the life-course (e.g., [25,26]), among other factors.In this paper, I will test for differences in friendship expectations, and its configurations, according to gender, age cohort, education and marital status.

Method
This article draws on data from a national survey about social attitudes and values in Portugal (n = 1142).It is important to stress that the main objective of this survey was not to collect data about friendship.Rather, two specific open questions about the topic were added to the end of the survey, minimizing the impact of the different nature of the subject covered on the answers, both to these specific questions, and to the ones that preceded them.The questions were "To you, what is a good friend?";"And [to you what is] an intimate friend?".The aim was to approach a maximum diversity of meanings that friendship may entail.Considering the nature of the data collection method (face-to-face survey, with a maximum length of 20 min), answers were predominantly short and concise, ranging from 1 single word (e.g., "trust", "support", or "family"), to 25 words, with an average of 7.2 and a median of 7 words ("good friend"); and from 1 word to 25 words, with an average of 5.5 and a median of 5 words ("intimate friend").These values are usual in surveys using similar methodologies.Moreover, although succinct, answers provided relevant information, given the ultimate aim of the study: to quantify the prevalence of the more salient meanings of friendship in a nationwide representative sample, as it is done in several surveys about social attitudes and social values.
The study and fieldwork were conducted by CESOP (Research Centre of Public Opinion of the Catholic University of Portugal), during the months of February and March 2014.In terms of sample distribution, 47% (n = 537) were men and 53% were women (n = 605).Age ranged from 18 to 96 years old (M = 49.34,SD = 17.98): 9% were between 18 and 24 years old, 15% were between 25 and 34, 19% were between 35 and 44, 17% were between 45 and 54, 16% were between 55 and 64 and finally 24% were 65 years old or more.In tandem with this age distribution, the sample also mirrors a reality where, despite huge efforts and achievements regarding education during the last four decades, levels of education in Portugal are still lower than the average of the EU or OECD countries [23].Though only 6% had no degree of education at all, 39% had only frequented school up to 6 years, 16% up to 9 years, 21% had a secondary education and 18% had a university degree.As for marital status, 27% were singles, 54% were married (of which 45% in church), 5% lived with a partner, 7% divorced and 8% widows.
Content analysis followed a semi-inductive methodology: on the one hand, a grounded-theory approach was privileged, as thematic categories were created bottom-up, keeping the very expressions used by respondents as much as possible; on the other hand, a set of categories was already settled, considering existing studies on the meanings of friendship [2,3].However, the analysis was particularly careful not to assume these studies' [2,3] findings as the main categories to code the material, for several reasons.Firstly, the subject of these studies was different and more specific.With Monsour [3] focusing on intimacy, his categories were especially helpful to understand the answers to my second question, but not to the first.As for Parks and Floyd's research [2], their interest was particularly on the differences between intimacy and closeness, whereas my focus was on disentangling the normative meanings of friendship in more general terms (a good friend) and then focusing on intimacy as well (an intimate friend).Moreover, while both these studies were conducted in a different societal context (USA) and based on non-probabilistic samples of university students, the present study is based on a nationally representative sample of a southern European country.Other studies on this topic, focusing on specific populations (gays and lesbians [27]; elderly [25]) also use small and non-representative samples, for the purpose of their subjects and hypothesis testing, which calls for special attention when it comes to comparing results.
Certain keywords were used as markers for coding: for instance, if an answer used the words "trust", or "rely on", or "count on", then it would be coded as trust.However, meaning always had primacy over the presence/absence of specific words: even when the word was absent, if the meaning pointed to the idea of "trust" then it would be coded under that category.As usual, many answers were coded under more than one category, as some individuals gave complex definitions of friendship.Special attention was given to overlapping categories and regular tabulation across the process enabled the recoding of categories in order to make them mutually exclusive, as much as possible.However, due to the very nature of the subject, it is not possible to avoid completely a certain overlap, which in itself gives us an account of the complexity and naturalized nature of social representations in this matter.An important decision was made regarding two of the most cited categories: presence and unconditional support.Monsour notes having coded under "unconditional support" answers that "indicated that intimacy meant "being there for one another through the good and bad times" and/or "showing support and concern for one another" ([3], p. 283).Instead, my results show a substantive difference in answers that used mainly or exclusively the idea of "being there" (without any necessary expectation of getting help), and answers that referred explicitly to "be able to count on", or referred to "need", "support", "help", among others.Even so, whenever definitions suggested both meanings, they were coded under both categories (n = 40).Moreover, and also differently from Monsour, a separate category for "concern and caring" was created, as well as for "sharing" (whenever it concerned sharing things, ideas, opinions, talk).
Categories were then statistically analysed using SPSS, firstly for descriptive purposes, and after using bivariate statistics to begin to draw a profile of the respondents, according to their definitions of friendship.After analysing each of the two questions, exploratory multivariate analysis was performed, in order to identify major groups of meanings given to friendship and, eventually, to build a typology of respondents according to those salient meanings.Firstly, using the multiple correspondence analysis (MCA) method, several analyses were run, using as active variables the most cited categories in each of the two questions: the cut off criteria was that only categories with over 10% of answers were retained.All variables were dichotomised, coded 1 (yes) and 2 (no).The exploratory analysis of the inertias, comprising six possible dimensions, enabled us to identify that the first two contributed most to the total explained variance (the first with 21.5%, the second with 20.5%).These were the ones chosen to perform the final analysis and observe both the discrimination measures and the way categories distributed themselves across the two-dimension plot.Discrimination measures also enabled the identification of the categories that contribute the most to each one of the two dimensions.
After identifying, through MCA, groups of answers (i.e., groups of dominant meanings given to friendship), a cluster analysis enabled the identification of groups of respondents.The procedure followed was to save the object scores of the MCA and use them to perform a K-Means cluster analysis with four clusters, a number inspired by the identified groups of categories disposed on the plan.These clusters were then projected as supplementary variables in the two-dimension plan of joint categories, confirming the consistency of the classification.The groups were then characterized, firstly by their constitutive original variables (the categories); and secondly by socio-demographic and structural variables (gender, age, education, marital status).

What is a "Good Friend"? Presence, Support and Trust
The content analysis of the open question "To you, what is a good friend?"revealed considerable variety, coded into 35 categories.The qualities of a good friend that stood out were unconditional presence ("being there in good and bad moments", 28%); unconditional support (being able to count on that person no matter the situation, in good and bad moments, 26%); and trust (18%).Other specific kinds of "presence" and "support" appeared, which were then merged into the same category: 7% of the respondents referred to simply "help" or "support", 8% referred to regular or daily presence, and less than 1% mentioned "being present even when distant", as well as being present particularly in good and happy moments in life.Though "unconditional presence" and "unconditional support" may easily be understood as overlapping categories, they also reveal an interesting distinction in terms of the norms and moral rules that people use to define a friend: between simple presence ("just being there") and effective support (providing help, being able to count on that person no matter the circumstances).
Other than these major categories, respondents identified an extensive array of meanings of a "good friend", ranging from moral qualities of the friend towards themselves (sincerity, loyalty, truthfulness, honesty or solidarity), to attitudes towards the friend himself (being a good listener, giving advice and guidance, being understanding, lack of moral judgment) and to more general moral qualities (being serious, humble or a good person).
Considering only the top three mentioned categories (presence, unconditional support and trust), we find major differences in these answers according to gender, age, education and marital status.Except for support (non-significant differences), women tend to choose, more than men, (co)presence as major meaning of friendship (28% against 23%, χ 2 = 3.44, p < 0.05), as well as trust (20% against 14%, χ 2 = 6.71, p < 0.01).However, age has a greater effect on the perceptions of a good friend: there is a significant difference between the youngest of the sample (18)(19)(20)(21)(22)(23)(24) and the oldest (65 and more), with the younger systematically identifying, more than the older, a good friend as someone who is always present, who provides unconditional support and whom one can trust one's life (p < 0.001).
This finding is not unrelated to the importance of education as a major predictor of differences in attitudes towards various issues.In fact, respondents with a higher education degree distinguish themselves widely from the rest, namely when it comes to identify (co)presence (being there daily, in good and bad moments, or whenever necessary) as a major characteristic of a good friend (40%, against 18% of respondents with basic school degree, χ 2 = 54.86,p < 0.001).The most educated ones are also the ones who define a good friend as someone who provides unconditional support.Interestingly, differences among educational levels are non-significant regarding trust.
Marital status also has an impact on respondents' imaginary of a good friend: those co-habiting without being married (either by church or civil ceremony), as well as singles, tend to define a good friend as someone who is present (χ 2 = 31.11,p < 0.001); singles also identify, more than other groups, a good friend as someone who provides unconditional support (χ 2 = 21.42,p < 0.01) and trust (χ 2 = 22.75, p < 0.001).

What about an "Intimate
Friend"? Self-Disclosure, Trust, Family Since this question followed the effort to define a "good friend", unsurprisingly 14% of the respondents answered that an intimate friend was "the same as a good friend".We might also interpret this either as a lack of distinction between the two dimensions of friendship, or as part of a common difficulty in defining something which is over-naturalized in private and public discourses.Confirming this assumption, 41 respondents (3%) were unable to provide any definition, or simply declared "I don't know", and 24 (2%) provided tautological definitions, such as "if he's intimate, it's because he's intimate", "if he's a friend, it's because he's intimate", or "a good friend with total intimacy".Answers that were unable to distinguish an intimate from a good friend do not differ significantly according to gender, age, educational level or marital status.
For those who accepted the challenge to define an "intimate friend", the results are interestingly different from those regarding a "good friend".The most cited category was self-disclosure (17%, n = 181), followed by trust (16%, n = 174) and family (14%, n = 144), equally placed with "the same as a good friend" (14%, n = 145).In contrast to what happened with the definition of a "good friend", "presence" (being always there, regularly and unconditionally, in good and bad moments) only gathers 5% of the answers; and "unconditional support" only gathers 7% of the answers.These results suggest that, for the vast majority of the respondents, a clear difference is made between both ways of defining friendship ("good friend" and "intimate friend").
The first most cited characteristic of intimate friendship was self-disclosure.Typical answers coded under this category are: "he's the person to whom I tell things that I don't tell to anyone else/not even to myself"; "he's a confidant"; "someone with whom I share all my secrets"; "someone with whom I can get things off my chest".As in Monsour's study [3], many of these answers included the word "sharing" (Monsour refers 30%, while 25% of the answers in the present study included either the word or its derivatives).
The second most cited characteristic of an intimate friendship in the present study was trust, which came only in fifth or fourth place in Monsour's study (according to gender and type of friendship, cross or same-sex), and in sixth place in Parks and Floyd's (though part of this category overlaps with self-disclosure, namely when it means "sharing or keeping secrets"; [2], p. 94).Trust seems therefore to be more vital to the respondents of the present study, aligned with the importance already given to it to define "a good friend".There is a slight difference in ranking: it came in third place in defining a "good friend" (with 17% of answers), and in second in defining an "intimate friend" (with 16% of answers).Examples of answers coded as trust are: "someone to whom we confide everything"; "someone to whom we confide as a child"; "person in which we can always rely on"; "someone I know I can fully trust".
The case of the category family is quite different.Its rise is worth noticing and demands further exploration.This category, which summed the total of 14% answers (n = 144), served to code all the answers that referred generally to "family" (n = 40), as well as to those that identified specific close relatives: husband, wife and partners (n = 45); sons, daughters and grandchildren (n = 26); parents and grandparents (n = 4); and brothers and sisters (n = 3).It is worth noticing that only very close kin were mentioned: other kinds of extended kin-such as uncles, aunts, nieces or nephews, cousins, as well as in-laws-were left outside this definition of intimacy.Within the category of family were also coded all the answers that identified an intimate friend "as family", or "almost family", "like a brother", "it's a brother" (n = 40).
Answers coded as self-disclosure and family show significant differences according to socio-demographic features of the respondents, though in opposite ways.There is a direct relationship between age and education and self-disclosure, with the youngest (χ 2 = 26.02,p < 0.001) and the most educated, with university degrees (χ 2 = 46.62,p < 0.001) predominantly identifying it as a major feature of intimate friendship.When it comes to family, we find an inverse relationship: it is the elderly (χ 2 = 38.99,p < 0.001) and the less educated (χ 2 = 19.09,p < 0.01) that tend to describe an intimate friend as (someone of the) family.

An Exploratory Typology of the Meanings of Friendship
Having unpacked the meanings of friendship in two dimensions (good and intimate friend), is it possible to join them in order to make sense of the dominant social norms that shape our expectations towards friendship?A multiple correspondence analysis of the answers to both questions outlined above (in Sections 3.1 and 3.2) enabled the identification of four main meanings ascribed to friendship by respondents, i.e., the most salient attribute in their mind when referring to a good and an intimate friend, according to two main dimensions (cf.Method).The first dimension discriminates mainly between considering "friends as family", as well as valuing trust within an intimate friendship relation; the second discriminates answers mainly between valuing (or not) trust, presence and unconditional support from good friends, as well as self-disclosure in intimate friendship.Based on the categories that contribute most to each dimension, we named dimension 1 as less institutionalized bonds oriented (valuing self-disclosure, trust, presence and unconditional support independently of the kinship, or kinship-like, bond); and dimension 2 more institutionalized bonds oriented (valuing husband, wife, sons, daughters, parents and grandparents, and family in general).
Figure 1 shows us how the most cited categories to describe both a good and an intimate friend distribute themselves across these two dimensions.It is possible to identify four major groups of answers, as well as four related groups of respondents (cf.Method).We can observe how the distribution of categories enables the construction of a typology of meanings given to friendship.Starting from the down-right side of the figure, and following a clockwise direction, four types or groups can be identified: (1) the family-oriented: a group for whom friendship is basically identified with or as family, at the same time that self-disclosure, presence and unconditional support are not chosen as determinant features (meaning that they do not immediately rise as the most prominent attributes in the mind of respondents); family is cited generally, but also identifying specific persons and bonds with an institutional dimension, such as husband/wife, sons and daughters, parents and grandparents; here we find also those that choose these institutionalized bonds to describe the intensity of their friendships ("like family"; "like a brother").This group is identified as cluster 1 and represents 40% of the sample.This high percentage does not refer exclusively to those who explicitly choose "family" (that are 14% of the total sample, and 31% of this specific cluster), but also to those who did not mention in their answers self-disclosure, unconditional support or presence as a major salient attribute of friendship.Typical responses within this cluster would be "my husband", "my wife", "my children and my husband", "my family", or "he's like a brother".(2) the trust-oriented: a group for whom friendship is mainly about trust, but in which respondents did not primarily choose unconditional support, presence or family; it is identified as cluster 2 and comprises 19% of the sample; typical answers within this cluster would be "someone to whom we trust everything", "someone I know I can always/completely trust"; "someone to whom I would trust my child".(3) the self-oriented: a group for whom friendship is mainly a synonym of self-disclosure and unconditional support, but who has not chosen as the most salient attribute of friendship presence, family or trust.The fact that it is possible to identify clearly two groups regarding presence and unconditional support corroborates the choice of not coding these two kinds of answers into the same category.It seems that there is a verifiable difference between counting on someone's help/support and expecting that person to be there, to be present.This happens despite the persistence of a certain overlap of both types (44% of the presence-oriented chose also unconditional support).This group is identified as cluster 3, and comprises 13% of the sample.It was named self-oriented as it describes a more self-centred approach to friendship, in which a friend is represented as someone who is at the service of Ego: providing appropriate conditions for his or her inner-life construction and display (self-understanding, self-narrative), as well as providing all kinds of instrumental and expressive support.Typical answers within this cluster would be "someone who knows as much about us as ourselves", "someone I tell everything/things I don't tell to anyone else", or "someone who supports us unconditionally".( 4) the presence-oriented: a group for whom friendship is basically a synonym of presence; this category comprises "being there" on a permanent, regular or even daily basis, but also in difficult moments, in good or happy moments, or "whenever necessary"; this group is identified as cluster 4, and represents 29% of the sample.Typical answers within this cluster would be "someone who is there in good and bad times" or "someone who is there whenever we need".
What is the profile of the respondents that belong to each one of these groups?As we can see in Table 1, the results of the bivariate analysis show a significant relationship between the belonging to each one of the clusters and age, education and marital status.Gender does not reveal significant differences.In cluster 1-Family-Oriented, we find mainly the older and less educated individuals, as well as the married ones, divorced and widows.Marital status may be interpreted as an indicator of values and practices regarding personal and intimate life.Therefore, it is not surprising that it is among those with more institutional/tradition-oriented bonds (namely related to the institution of marriage) that we find mainly Family-Oriented representations of friendship.This result is also correlated with the age of the respondent, as the older are more likely to be widows.In cluster 2-Trust-Oriented, we find mainly the younger (until 34) and singles.In cluster 3-Self-Oriented, we find also the younger (until 34) and the more educated (namely those with a secondary or university education).However, neither in cluster 2 nor in cluster 3 do we find very pronounced differences, in terms of structural predictors.Finally, in cluster 4-Presence-Oriented, we find mainly the younger (until 44) and educated (secondary or university education), as well as those living with a partner (no institutionalized union), singles, and divorced/separated.In contrast to cluster 1 (Family-Oriented), it is among those whose intimate life assumes non-institutionalized forms that we find mainly a presence-oriented representation of friendship.
Because these structural predictors are, expectedly, interrelated, logistic regressions were run, in order to better observe the explanatory power of each predictor, while controlling for the effects of all other variables in the model.In these logistic regression analyses, marital status loses its relevance to explain group membership, suggesting that differences among singles and married, or divorced, or widowed, may instead be associated with the stage of the life-course that, according to age, individuals go through; or alternatively, to a set of values enhanced by education.Results are shown in Table 2.As we can see, membership to cluster 1 (Family-Oriented) is by far the most well explained by the chosen variables.Age, articulated with level of education, is a strong predictor of membership to family-oriented representations of friendship, the odds of belonging to this type rising among the elderly, when compared to the cohort 18-25, and being a lot higher among the oldest cohort (65 and more).The same can be said about education, the probability of belonging to this cluster being higher among the less educated, when compared to university level, namely those with basic education; no significant differences are shown between the secondary and university level.Belonging to the other three clusters is harder to predict, at least with the available variables in the model.Membership to cluster 2 (Trust-Oriented) is only weakly explained by age, the probability of sharing this kind of representation of friendship being lower among the cohort 35-44 (when compared to [18][19][20][21][22][23][24][25].Membership to Cluster 3 (Self-Oriented) is partially explained by education.Those with the lower levels of education show a lower probability of belonging to this group.As for cluster 4 (Presence-Oriented), it is partially explained by age and education: being 65 years old or more, and having a basic education degree (less than 9 years in school) decreases the odds of belonging to this group and, therefore, sharing a representation according to which friendship is basically about "being there" in ordinary as well as in exceptional circumstances.

Discussion
The results suggest that, while certain aspects of friendship norms appear equally important in the studied context (Portugal)-here taken as a case study-confirming its hegemony across different cultures, other aspects reveal differences that call our attention to how the specificities of different societal backgrounds may bring diversity into existing definitions of friendship.Moreover, some aspects that have systematically been identified as important to define the ideal friend have not shown to be relevant in this specific cultural setting.
Focusing firstly on similarities, results regarding the definition of a good friend are in line with previous studies, as they underline the relevance of (unconditional) help, support and trust as major dimensions of friendship expectations.In their study of 270 university students, Parks and Floyd ([2], p. 94) found that help and support was the second most cited characteristic of a close relationship (37% of answers), while trust came in sixth place (20% of answers).
In fact, an important result of the present study that also corroborates previous research is the overarching importance of trust.In the present study, this might be explained by the particular configuration of inter-personal trust in Portugal, with all its implications for (lack of) social capital.Portugal has one of the lowest levels of interpersonal trust among European countries, with only 12.3% of the respondents giving a positive answer to the question "Most people can be trusted", occupying almost the last place in rank (only above Romania), way below the participant countries average (30.5%) and even more of Scandinavian countries, like Sweden or Denmark, in which averages rise to more than 66% [21,28].This context of a general lack of interpersonal trust as a structural and enduring characteristic of a country with many inequalities in the way it entered in late modernity may partially explain why there are no significant differences in the answers coded as trust according to gender, age, education, marital status or region.Trust seems to be, unsurprisingly, a cross-cutting feature attributed to friendship-an asset that is as much treasured as it lacks as a structuring social value.
The present study also corroborates the importance of the life-course in the expectations we develop and maintain towards friends.Results show that the marital status has an impact in the representation of a good friend, with those co-habiting without being married, as well as singles, defining a friend as someone who is present, and singles electing unconditional support and trust.This goes in line with Wall and Gouveia's [19] study about changing family meanings in Portugal, in which friendship-oriented configurations (mainly composed by friends and parents) were more likely to be found among the younger, the odds of belonging to this kind of family configuration decreasing with partnership, parenthood and becoming a widow.
Results also showed a partial convergence with already existing studies regarding the definition of an intimate friend, with self-disclosure emerging as a major feature of friendship.The capacity to trust and confide in the other as one of the major rules of friendship was first underlined by Argyle and Henderson [29], and after corroborated by others, such as Monsour [3], Parks and Floyd [2] or Fehr [24].The replication of this result in the present study is even more interesting given the methodological differences that separate it from the cited ones.It is therefore worth noticing that, despite all differences, self-disclosure still represents the most cited meaning of intimacy within friendship, a result that points to its pervasiveness as a cross-national and cross-cultural characteristic of intimacy and friendship.Friendships, and particularly intimate ones, seem to fit, and at the same time contribute to the contemporary occidental "rituals of confession" that, according to Foucault [30], have given form to the construction of the Self since the 18th and 19th centuries.Corroborating this, Parks and Floyd show how lack or limited self-disclosure is referred to by their respondents as the second cause for indicating that the friendship is not intimate ([2], p. 97).
Focusing now on the specificities and differences of the present study, maybe the most important and interesting result is the prevalence of family (kinship ties) in the definition of an intimate friend.On the one hand, since it is not possible to know what respondents meant by it, the subjective nature of such a response remains blurred and thus open to further enquiry through qualitative research.On the other hand, the fact that for many respondents an "intimate friend" is "my wife/husband", "my children", or simply "my family", stands in itself interesting.To better understand this result one must put it in the context of previous research about family in Portugal, where values, attitudes and practices regarding family life reflect a paradoxical societal context marked by accelerated structural changes that coexist with persistent inequalities (in a process that some authors have called "unfinished modernity" [20]).
In terms of values, for the Portuguese, as for many other European, "family" rises as the most important aspect in life ( [31], p. 50).Despite the increasing rates of births outside marriage, the majority of the Portuguese believe that being in a coupleship is a condition of happiness, while only less than a quarter think that marriage is an old-fashioned institution ( [31], pp.52-53).Moreover, blood ties are seen as implying a universal code of mutual and unconditional duties and obligations between parents and children ( [31], p. 83).Such persistence of the importance of kinship can also be found regarding family meanings and practices.Wall and Gouveia [19] identified four major types of family configurations: nuclear (46%), friendship oriented (24%), beanpole (16%) and sibling oriented (14%).The authors show how networks of significant family ties in Portugal highlight the "still fundamental generating principles of partnership, consanguinity and co-residency in building up family bonds" ( [19], p. 369).
Therefore, this outcome suggests a two-folded interpretation.On the one hand, the great majority of the answers coded under this category refer to institutional ties, either of blood or alliance (children, parents, husband, wife), and not "family-like" ties (e.g., friends "like brothers").This confirms Wall and Gouveia's [19] conclusions about "the overarching salience of kinship" in individuals' networks of family ties in Portuguese society, with a numerical predominance of the nuclear configuration (couple and children, 46%), showing the country's specific normative and cultural context in which family obligations have historically shaped the norms of private life.
On the other hand, it may be interpreted as an indicator of patterns of suffusion (instead of mutual exclusion between the categories of "friends" and "family"), whenever participants identified an intimate friend "as family", or "almost family", "like a brother", "it's a brother" (n = 40).Here we have a different version of the same "family oriented" representation of friendship: one in which kinship ties are major referents, but without necessarily implying blood or alliance bonds.Recent research about Portugal also shows that friends represent 5% of alters named as meaningful ties, being cited by 13% of the respondents.Moreover, 7% of the persons perceived as family are non-kin [19].
Either way, this representation of (intimate) friendship as family confirms the processes that Pahl and Spencer [9,10] called suffusion, of blurring boundaries between kin and non-kin, with roles among members of the personal community being inter-changeable: friends can be "like family"; and family can be "like friends".If the latter ("family-like friends") is also present, it is not, however, the most prevalent: strong kinship bonds are predominantly represented as the ultimate form of intimacy that one can experience within friendship.For some, husbands, wives, fathers, mothers, sons, daughters, are the most accomplished definition of friendship.
Maybe this explains in part another interesting difference between these results and previous research: the importance given to physical and sexual contact.In Monsour's study ( [3], p. 286), physical contact comes in fourth place, in a set of seven ascribed meanings of intimacy; and sexual contact in seventh.As for Parks and Floyd's study ( [2], p. 97), 50% of the respondents felt that the main reason for not considering a friendship as intimate was the fact that intimacy was seen as a component of a romantic and/or sexual relationship, and not of a friendship.In the present study, however, only two respondents made indirect references to physical or sexual contact ("it is the kind of intimacy you don't have with friends").It might be argued that, when respondents named husbands, wives, boyfriends and girlfriends, they were also indirectly making a reference to physical or sexual contact as a component of intimacy within friendship, but there is no way to unpack those meanings by the words they used in this study.
Other interesting differences can be found in definitions of friendship that go back to the dawn of modernity.Reciprocity, which is seen as a major moral feature of friendship in modernity [32], appears only embedded in other categories, in answers such as "helping each other when necessary", "loving each other", "caring for each other".This may point to its over-naturalization, and consequent assimilation in other representations (help, support, affect, care).Another central characteristic of modern friendship is its supposedly altruistic nature [32].However, this was only directly referred to by 27 (2.3%)respondents.Care and affect did not gather many responses, suggesting that either the affective nature of friendships is so naturalized that respondents do not feel the need to mention it, or that other instrumental dimensions are seen as more distinctive and/or important.Another interesting result is that only 14 respondents (less than 1%) answered that they had no friends, or that "there is no such thing as a good friend", or that "a good friend is the most difficult thing to find".Thus, it seems that cynical views on the nature of human relations, namely of friendship, that go back to pre-modern conceptions, before the 18th century [32], are a clear minority in this study.
Regarding the four types of friendship norms or expectations revealed by multivariate analysis, two main things must be stressed.Firstly, they show configurations that are different from previous typologies, drawn either from open-questions or item testing analysis, namely in the importance that is given to kinship ties to define a friend.This might be interpreted as a singularity of the social context in which the data were collected, characterized by particular tensions regarding personal life and a strong ideological commitment to the family.Such a result calls for including different societal contexts in the analysis of friendship and its cultural manifestations, therefore adding new knowledge to the subject.Secondly, these four types of friendship norms are not universal within the same social context, with age, education and marital status influencing the odds of belonging to each one of them.
Family-Oriented representations of friendship are mainly shared by the older and less educated, showing a clear prevalence of tradition.Here, suffusion is not primarily a synonym of modern values (for instance of new forms of intimacy, in which friends are seen as family), but rather of the persistence of pre-modern characteristics, namely of the importance of kinship relations as a means of social organization to make social ties stable across time and space [33].
As for Trust-Oriented representations of friendship, the fact that they are difficult to predict-with only age (35)(36)(37)(38)(39)(40)(41)(42)(43)(44) showing a significant influence in the odds of belonging to this group-suggests, as it has already been argued, that trust is a cross-cutting feature ascribed to friendship in a society in which levels of interpersonal trust are very low.In the complex process of transition from pre-modernity to modernity, in which the opposite of friend is not anymore "enemy", but acquaintances or co-workers, friendship emerges as an ultimate (idealized) resource of interpersonal trust, which becomes, on a personal level, a common project "to be 'worked at' by the parties involved, and demands the opening out of the individual to the other" ( [33], p. 121).This hegemonic importance of trust also points to the importance of friendship as a means of accessing a sense of ontological security, in modern societies.Friends may well assume a central role in reassuring the modern individual regarding the continuity of their self-identity, reaffirming a "sense of the reliability of persons and things, so central to the notion of trust" ( [33], p. 92).Confirming this, and inverting the perspective, Smart et al. [11] have shown how difficult friendships install in the individual a feeling of existential anxiety and ontological insecurity, because they seriously, and sometimes permanently, jeopardize interpersonal trust (this state of existential anxiety and fear being the precise opposite of trust [33]).
Belonging to the other two clusters is better understood in relation to what has been described as major characteristics of modernity, in contrast with pre-modern or traditional societies [32,33].As described above, cluster 3 (Self-Oriented) comprises those for whom friendship is mainly a synonym of self-disclosure and unconditional support.As Giddens [33] and others [30,34,35] have argued, self-disclosure and authenticity are major components of the modern individual: who must make of his/her life his/her own work of art, discovering and revealing his/her "true nature and inner self" (being authentic).Significant others are determinant to confirm this reflexive identity, abiding by the rules of affect and not of institutionalized bonds.Furthermore, access to information, and having the means to process and incorporate it, in a reflexive process, is decisive to the constitution of this "modern individual", and of modernity itself.Therefore, while education remains a scarce resource, while "literacy is the monopoly of the few, the routinisation of daily life remains bound up with tradition in the old sense" ( [33], p. 38).The same line of arguments can be used to understand membership to cluster 4, Presence-Oriented.

Conclusions
One limitation of this study is the fact that it draws on data provided only by two open questions of a survey.This is a consequence of the quantitative methodology that was used, a survey applied to a nationally representative sample (n = 1142), where open ended questions are usually restricted, as it is the case of large scale surveys regarding social attitudes and values.Rather than seeking a contextual and complex production of discourses, this methodology seeks to capture the salience of certain categories for the respondents, as well as its diversity.As it was expected, the two questions did not provide long discourses, but rather succinct and concise answers, which nevertheless gave new and relevant information to meet the ultimate goal of the study: to quantify the prevalence of the more salient meanings of friendship in a nationwide representative sample.In order to overcome these limitations, research proceeded with ethnography and qualitative interviews, which will provide in-depth and contextual information to complement the results that are discussed in this paper.
The most important result of this study is the construction of a new typology of friendship meanings, with four types: family-oriented; trust-oriented; self-oriented; and presence-oriented, each one of which may be referred to a group (cluster) of respondents, with a particular social profile.This typology highlights the complexity of configurations of friendship representations, but one that can nevertheless be understood as a bipolar reality, with two apparently contradictory movements.On the one hand, there is a more tradition-oriented set of representations, in which kinship ties have the primacy in defining friendship, confirming the non-linearity of processes of individualization and highlighting, instead, their discontinuities.On the other hand, there is a more individualized set of representations, related to the construction and maintenance of the Self (self-disclosure; unconditional support and presence), in line with the literature that has highlighted the convergence to reflexive and self-centred practices and values, in western societies, as a result of the individualization process.These contradictory tendencies are unequally distributed in terms of age cohorts and education, with tradition-oriented representations being mainly shared by the elderly and less educated and, conversely, more individualized orientations shared by the younger and more educated.
This means, on the one hand, that friendship meanings (and practices) are unequally distributed according to different sets of capitals and resources.On the other hand, friendship may itself enhance (or limit) these resources, contributing to diminishing (or aggravating) such inequalities.This highlights the importance of the specific social context in which research about friendship takes place.In the case discussed in this article, relevant contextual factors include demographic characteristics (an ageing country, with a high and growing percentage of individuals over 65, and a low fertility rate, accounting for demographic challenges shared by other countries in Europe); structural factors (education levels, gender equality) and cultural ones (values and practices regarding family and personal life).The unequal distribution of representations of friendship across the social fabric thus reveal complex and multi-layered ways of transition to late-modernity and more individualized societies, and the modes they interfere with personal life and the economy of affects.In this sense, the Portuguese case helps to shed some light onto the ways personal relationships-and namely friendships-are built in contemporary western societies, contesting assumptions of linear evolution towards individualization, conceived as one-dimensional processes.However, one limitation of this study is the lack of representative and equivalent data that would enable a comparative international analysis of the subject (friendship norms and expectations).In order to further explore the possible communalities and differences between southern European societies, such as the Portuguese, and other European and world cultures, additional comparative research is necessary.

Table 1 .
Belonging to type of friendship representations (cluster membership), by structural predictors (crosstabs).