ai Intemational Journal of Bifurcation and Chaos, Vol. 7, No. 11 ( © World Scientific Publishing Company 1997) 2611-2619 CYCLIC DYNAMICS IN ROMANTIC RELATIONSHIPS ALESSANDRA GRAGNANI* Dipartimento di Elettronica ¢ Informazione, Politecnico di Milano, Via Ponzio 34/5, 20133 Milano, Italy SERGIO RINALDIt Dipartimento di Elettronica e Informazione, Politecnico di Milano, Via Ponzio 34/5, 20133 Milano, Italy , GUSTAV FEICHTINGER' Institut fir Okonometrie, Operations Research und Systemtheorie, Technische Universitat Wien, Argentinierstrasse 8/119, A 1040 Wien, Austria Received July 17, 19: Revised January 23, 1997 ‘Minimal models composed of two ordinary differential equations are considered in this paper to mimic the dynamics of the feelings between two persons. In accordance with attachment theory, individuals are divided into secure and non-secure individuals, and synergic and non-synergi¢ individuals, for a total of four different classes. Then, it is shown that couples composed of secure individuals, as well as those composed of non-synergic individuals can only have stationary modes of behavior. By contrast, couples composed of a secure and synergic individual and a non-secure and non-synergic individual can experience cyclic dynamics. In other words, the coexistence of insecurity and synergism in the couple is the minimum ingredient for cyclic love dynamics. The result is obtained through a detailed local and global bifurcation analysis of the model. Supercritical Hopf, fold and homoclinic bifurcation curves are numerically detected around 8 Bogdanov-Takens codimension-2 bifurcation point. The existence of a codimension- 2 homoclinic bifurcation is also ascertained. The bifurcation structure allows one to identify the role played by individual synergism and reactiveness to partners love and appeal. It also explains why ageing has a stabilizing effect on the dynamics of the feelings. All results are in agreement with common wisdom on the argument. Possible extensions are briefly discussed at the end of the paper. 1. Introduction ‘This paper deals with love dynamics, a subject which has received remarkable attention in the last few years. The problem falls in the field of so- cial psychology, where interpersonal relationships are the topic of major concern. Romantic relation- ships are somehow the most simple case since they involve only two individuals [Sternberg & Baines, “Esmailigragnani@elet_polimi.it *E-mailinaldiGelet.polimi.it ‘E-mail: or@e119wst.tuwien.acat 2611 1988]. The most common approach to the problem is rooted in attachment theory [Bowlby, 1969, 1973, 1980] which explains why infants become emotion- ally attached to their primary caregivers and why they often experience emotional distress when phys- ically separated from them. Empirical research hhas focused on different attachment styles (se- cure, anxious, avoidant, ...) in children [Ainsworth et al., 1978] and adult individuals (Hazan & Shaver,2612 A. Gragnani et al 1987; Collins & Read, 1990; Feeney & Noller, 1990; Bartholomew & Horowitz, 1991; Griffin & Bartholomew, 1994] and several hypotheses have been generated about the nature and emotional quality of romantic relationships possessed by peo- ple who exhibit different attachment styles. In this context, the attachment style of an individual is be- lieved to persist relatively unchanged all throughout his life, even if a recent study has pointed out some exceptions [Fuller & Fincham, 1995]. In conclusion, one can reasonably argue that the main character- istics of love-stories should be largely dictated by the attachment styles of the individuals involved. Love-stories are dynamic processes that start from zero (two persons are completely indiffer- ent to each other when they first meet), develop (more or less quickly) and end up into some sort of regime. Real-life observations tell us that most times transients develop very regularly and asymp- totic regimes are stationary and associated to posi- tive romantic relationships. But there are also love- stories initially characterized by stormy patterns of feelings as well as cyclic regimes, like those identi- fied by Jones [1995] in Petrarch’s Canzoniere, the most celebrated book of love poems of the West ern world. The existence of “cycles” with different frequencies in romantic relationships has been rec- ognized in many experimental works. For example, there are easily detectable 24-hours oscillations in the feelings between two persons if at least one of them has pronounced daily variations of blood pres- sure. Weekly cycles have also been observed and are obviously due to the difference between weekends and working days. Moon cycles can be also present in young couples because the sexual activities are periodically inhibited by menstruation. All these cycles are high-frequency cycles which are forced by exogenous factors. On the contrary, the cycles we talk about in this paper are low-frequency and endogenously produced. From the phenomena de- scribed in the model one should expect cycles with periods ranging from a few months to a few years. For example, the love cycle discovered by Jones in Petrarch’s Canzoniere is about 4 years long. What has been said about steady and cyclic romantic re- lationships reminds very much the behavior of dy- namical systems tending toward equilibria or limit cycles. But observations also point out the exis- tence of multiple attractors. For example, it is known that steady and high quality romantic re- lationships can turn into a state of permanent an- tagonism after a disturbance, e.g. after a temporary infatuation of one of the two partners for another person. Finally, even bifurcations can be naturally invoked if one looks at the effects of age, which is a slowly varying parameter capable of transforming ‘tempestuous relationships into steady ones. ‘The above remarks suggest the use of differ- ential equations for modeling the dynamics of the feelings between two individuals. It is therefore not surprising if a few contributions have recently appeared along this line. They are rooted in a one-page pioneering paper by Strogatz [1988] enti- ‘led “love affairs and differential equations” and are briefly reviewed in the next section. Then, in Sec. 3 the mechanisms generating cyclical love dynamics are investigated through bifurcation analysis. The key point is the detection of a Bogdanov-Takens bifurcation and the result is that the simplest cou- ple with cyclic dynamics is composed of a secure and synergic individual and a non-secure and non- synergic individual (synergic individuals are those who increase their reactions to partner’s love and appeal when they are in love). Interpretation of the results and suggestions for further research conclude the paper. 2. Review of Previous Models ‘The models proposed up to now are minimal mod- els, in the sense that they have the lowest possi- ble number of state variables, namely one for each member of the couple. Such variables, indicated by 2 and zz, are a measure of the love of individual 1 and 2 for the partner. Positive values of x rep- resent positive feelings, ranging ftom friendship to passion, while negative values are associated with antagonism and disdain. Complete indifference is identified by ‘Minimal models are a crude simplification of re- ality. Firstly, because love is a complex mixture of different feelings (esteem, friendship, sexual satis- faction, ...) and can be hardly captured by a single variable. Secondly, because the tensions and emo- tions involved in the social life of a person cannot be included in a minimal model. In other words, only the interactions between the two individuals are taken into account, while the rest of the world is kept frozen and does not participate explicitly in the formation of love dynamics. This means that rather than attempting to be complete, the aim is to check which part of the behaviors observed in real life can in theory be explained by the few in- gredients included in the model.‘Three basic processes, namely oblivion, return and instinct, are assumed to be responsible of love dynamics. More precisely, the instantaneous rate of change ¢; of individual’s i love is assumed to be composed of three terms, i.e. B= O+R+h where the functions O;, R; and J; have different specifications in each model. Tn the first mode! discussed in the literature [Strogatz, 1988] (see also [Radzicki, 1993; Strogatz, 1994]) oblivion and instinct are neglected, while re- turn Rj, which interprets the reaction of individ- ual i to the partner's love 23, j # i, is assumed to be proportional to 2,. But the two proportional- ity coefficients have opposite sign, i.e. one of the ‘two lovers is a bit masochist and hates to be loved and loves to be hated. Thus, the model turns out to be a linear oscillator. Strogatz himself explains why he has made these very extreme assumptions: Hiis goal was to teach harmonic oscillators using “a topic that is already on the minds of many college students: The time evolution of a love-affair”. In the two other models oblivion and instinct are also present. Oblivion is specified as Oy = - a4; so that in the extreme case of an individual i who has lost the partner (Ri = I; = 0), 4 vanishes expo- nentially at a rate aj. For this reason, a; is called forgetting coefficient. Instinct J; describes the re- action of individual i to the partner's appeal A, Of course, it must be understood that appeal is, not mere physical attractiveness, but more prop- etly and in accordance with evolutionary theory, a suitable combination of different attributes among which age, education, earning potential and social position. Moreover, there might be gender differ- ‘ences in the relative weights of the combination [Feingold, 1990; Sprecher et al., 1994). In the second model (Rinaldi, 1996a] all pro- cesses are assumed to be linear and given by O=-oe R=Bay l= vA; with a;, 6; and 7%, i = 1,2, positive. Thus, the model turns out to be a positive linear system {Luenberger, 1979; Berman et al., 1989; Rinaldi & Farina, 1995] enjoying, as such, a number of re- markable properties. In particular, if the geometric mean Teactiveness to love (V/iB2) is smaller than Cyclic Dynamics in Romantic Relationships 2613 the geometric mean forgetting coefficient (,/@7@3) there is a unique positive and stable equilibrium Et = (c}, cf), and the two persons, completely indifferent to each other when they first meet, develop a love story characterized by increasing feel- ings. Moreover, the quality (}, z}) of the roman- tic relationship at equilibrium improves with the re- activeness to love (6;) and appeal (7:). Finally, an increase of the appeal A; of individual i gives rise to an increase of the feelings (z}, 2) of both in- dividuals at equilibrium, but the relative improve- ment is higher for the partner of individual é (in other words, there is a touch of altruism in a woman [man] who tries to improve her (his] appeal). Some of these individual properties can be used to infer properties at the community level: The main result along this line is that a community composed of NV linear couples is stable if and only if the partner of the nth most attractive woman of the community is the nth most attractive man (n = 1, 2,...,.N). This means that individual appeal is the driving force that creates order in our societies. In the third model [Rinaldi & Gragnani, 1996] return R; and instinct J; are still assumed to de- pend only upon 2; and Aj, respectively, but the dependence is nonlinear: Tt takes into account the traits of so-called secure individuals, who are the majority of the individuals in human populations. Secure individuals have positive mental models of themselves and of others and their romantic rela- tionships are characterized by intimacy, closeness, mutual respect and involvement. They react posi- tively to the partner's love and ate not afraid about someone becoming emotionally close to them. In conclusion, their return R; is an increasing func- tion of zj. Figure 1(a) shows the graph of a typi- cal return function of a secure individual (the same graph characterizes the function 1;(4,)). Note that RY and Ry denote the return for very large posi- tive and negative partner's feelings. Boundedness is a property that holds also for non-secure individ- uals: It interprets the psycho-physical mechanisms that prevent people from reaching dangerously high stresses. By contrast, the fact that the function in- creases with 2; is typical of secure individuals, since non-secure individuals react negatively to too high pressures and involvement [Griffin & Bartholomew, 1994], as shown in Fig. 1(b). The model is therefore (1)2614 A. Gragnani et al. @) Fig. 1. () Fig. 2. &) Return funetion Ri(zj) of secure individuals (a) and non-secure individuals (b). (b) State portraits of couples composed of secure individuals: Robust couples (a) and fragile couples (b). In (b) the boundary of the two basins of attraction (dashed line) is the stable manifold of a saddle. where the graphs of the functions R* and J” are like in Fig. 1(a). This model retains many of the properties of the simpler linear model, like the exis- tence of a stable positive equilibrium E*, and allows one to derive the same conclusions on the role of the appeals at community level. The main difference is that the nonlinearities can give rise to a negative stable equilibrium £7 (generated through a fold bi- furcation). Couples with a unique attractor (E*) are called robust, while couples with two attractors (E* and E-) are called fragile. Figure 2 shows the two corresponding state portraits. Note that frag- ile couples can switch from E* to E~ if one of the two individuals has a sudden drop in interest for the partner [see trajectories starting from points 1 and 2 in Fig, 2(b)]. ‘Before moving to the next section, we like to stress that limit cycles cannot exist in model (1) even if individuals are non-secure, i.e. even if the functions R* are like in Fig. 1(b). In fact, the di- vergence of the system (equal to — (ay +.a2)) does not change sign and Bendixon’s criterion implies the non-existence of limit cycles. This means that in a couple with cyclical love dynamics at least one indi- vidual i must have an instinct function J; depending also upon a; or a return function R; depending upon both state variables. This is related with synergism as discussed in the next section.3. Synergism and Cyclic Dynamics It is known that individual reactions can be en- hanced by love. For example, mothers have of- ten a biased view of the beauty of their children. This kind of phenomenon, here called synergism, has been empirically observed in a study on per- ception of physical attractiveness [Simpson et al., 1990] by comparing individuals involved in dating relationships with individuals not involved in them. Altbough we are not aware of any study pointing out the existence of synergism in the reaction to the partner's love, we can reasonably assume that also return functions can be enhanced by love. Thus, we consider reaction and instinct functions of the form R= (14 SR(2)) Res) @) Ke= (1+ SU(a)) (As) (3) where the functions Rj and J; are, by definition, the reactions of a non-synergic individual and the functions S*(z;) and S} (zi) are zero for z; < 0 and increasing, convex/concave and bounded for 2 > 0, as shown in Fig. 3. The upper bounds of the func- tions SF and S! are indicated by sf and sf and are called synergism coefficients. ‘We have shown in the preceding section that couples composed of non-synergic individuals, i.e., couples with s¥ = s! = 0, cannot have cyclic be- havior. Thus, synergism is necessary for generating cyclic love dynamics. It can be shown however, that synergism is not sufficient if the couple is composed of secure individuals [i. individuals characterized by reaction and instinct functions R* and I* like in Fig. 1(a)]. We prove this result by referring to the case of synergic instinct functions (3), ie. by 0 Xi ‘The graph of a typical synergism function Fig. 3 Cyclic Dynamics in Romantic Relationships 2615 analyzing the model 0421 + Ri(z2) + (1+ S{(e1))4(A2) te + Ry(x1) + (1+ 54(x2))15(A1) ; @ a2 A similar proof holds for the case of synergic return functions (2). The proof of non-existence of limit cycles in model (4) is as follows. Note first that the isocline = 0 can be given the form zz = z2(z1), be- cause the function Rj(z2) is invertible in the case of a secure individual see Fig. 1(a)]. Such isocline is composed by one, two or three different func- tions defined over disjoint intervals, because the inequality RY Sanz ~ (1+ S{(21))h (Az) < Rit is satisfied in one, two or three disjoint intervals (note that the function between the two inequal- ity signs is continuous and stationary at most two points). Obviously, the same properties hold for the isocline 42 = 0 which can be given the form 2) = 21(22). Figure 4 is a sketch of the isoclines in the case the first one (1 = 0) is defined on two in- tervals and the second (2 = 0) on a single interval. There are seven equilibria, which are either saddles (5) or nodes (Y) (a focus cannot exist in system (4) because the product of the elements on the antidiag- onal of the Jacobian matrix (dRj/dza) - (dR3/dz) is always positive). The nature of the seven equi- libria, can be immediately detected looking at the direction of the trajectories on the isoclines. Also invariant sets delimited by the isoclines can be eas- ily identified, as indicated in Fig. 4 (see shaded re- gions). Since, by construction, all equilibria are on the boundaries of these invariant sets, cycles cannot exist inside these sets. Thus, eventually, they must lie entirely outside, namely in the white regions of Fig. 4. But this is impossible because the union of the invariant sets is connected and expands to in- finity (note that isocline 2, = 0 [22 = 0] tends to infinity in the z2 [x1] direction). The proof of the non-existence of limit cycles given here for the case depicted in Fig. 4 (where the first isocline is defined ‘on two intervals and the second on one interval) can be repeated for all other cases of concern. This formally proves that secure individuals cannot have cyclic love dynamics even if they are synergic. In order to identify a case of cyclic dynamics we now consider couples composed of a secure and syn- ergic individual and a non-secure and non-synergic2616 A. Gragnani et al. Fig. 4. The isoclines (é1 =0, #2 = 0) and the equilibria of system (4) (N = node, 3 = saddle). The shaded regions are invariant sets. individual. Thus, the model is dy = ~anz1 + Ri(z2) + (1 + Sf (21) (An) fq = — 02% + R3(zi) + 13(Ai) i erences ey (6) (n= (aa -1) " Ritea/(1+ 22) vy _ f2mla= 21) ae (ee aya si(a) = He ‘ stat (+24) Figure 5 shows the bifurcation curves of model (5) and (6) in the two-dimensional space of synergism coefficient (sf) and maximum return (Rj*) of the first individual (a secure and synergic individual). ‘The organizing center is the Bogdanov-Takens codimension-2 bifurcation point BT in which three bifurcation curves merge: A supercritical Hopf (h), a fold (f) and a homoclinic (p). There are two other codimension-2 points: The cusp C’ and the degen- erate homoclinic bifurcation point D where a fold f and a homoclinic p merge giving rise to a saddle- node homoclinic bifurcation curve g. As a result there are five possible dynamic behaviors identified as 1, 2,..., 5. and described with a sketch of the cor- responding state portrait. In regions 1, 2, 3 the at- tractor is unique (an equilibrium in regions 1 and 2 where Rj and Rj are like in Figs. 1(a) and 1(b), re- spectively, and SJ is like in Fig. 3. For S{(z;) = 0 this model degenerates into model (1), which has no cycles. Hence, we should not expect cycles in model (6) for low values of the synergism coefficient. In order to prove that model (5) can have cyclic dynamics, we have performed a numerical but rather systematic analysis of its local and global bifurcations. For local bifurcations (Hopf and fold bifurcations) we have used LOCBIF, a pro- fessional software package for continuous-time dy- namical systems based on a continuation technique [Khibnik et al., 1993}. For global bifurcations (ho- moclinic bifurcations) we have used AUTO86 and we have followed a two-steps approach. First a homotopic method has been used to generate one point of the bifurcation curve and the correspond- ing homoclinic orbit. Then, a projection boundary condition method has been used to produce the en- tire homoclinic bifurcation curve through continua- tion, Both methods are described in [Champneys & Kuznetsov, 1994]. The analysis has been performed for various functional forms of Rj, Rj and S{ and the results have been consistent. The bifurcation portraits shown in the following make reference to the functional forms n<0 m20 n<0 (6) -A)/A+a)) 20 a <0 20 ee and a limit cycle in region 3), while in regions 4 and 5 there are alternative attractors. The dashed curve does not involve bifurcations of attractors, while all others do. The diagram confirms our expectations: There are no cycles if the synergism coefficient is low but there are cycles if individual 1 is highly sensitive (high reaction to partner's love Ry*+ and high synergism coefficient s/). ‘The same bifurcation structure has been de- tected varying other parameters. Figure 6 shows, for example, the bifurcations with respect to the synergism coefficient of the first individual and the appeal of the second. Again the Bogdanov-Takens bifurcation point is the organizing center.3 3 maximum return of secure individual R, 3.05 3.10 34s 320 ‘synergism coefficient of secure individual s,! 05, Fig. 5. lowing parameter setting: a1 = 0.2, o2 = 0.1, [i(4a) = I(x) = 0.95, = J cm 8 jo qm va 1s 18 synergism of secure individuals! Fig. 6. Bifurcation portrait of model (5) and (6) for the fol- lowing parameter setting: a; = 0.2, a2 = O.1, Ia(As) = 0.5, Rit = 005, Ry~ = ~0.06. ‘We have also analyzed the case of synergic re- turn functions, namely the model 121 + (1 + SP(21))Rj(22) + F7(42) © 99 + R3(x1) + 13(A1) a= f= Cyclic Dynamics in Romantic Relationships 2617 0.096 5 ‘maximum return of secure individual R, 200205 2102S synergism coefficient of secure individual s,* Fig. 7, Bifurcation portrait of model (7) for the following parameter setting: a, = 0.2, a2 = 0.1, Ji(Aa) = 0.05, 14(A1) = 0.05, Ri = — Rit. and we have, once more, obtained the same bifur- cation structure. Figure 7 shows, for example, the bifurcation diagram analogous to that of Fig. 5 for the functions Rj(z2), R3(z) and SP(x,) given by Eq. (6) with s{ replaced by sf. In conclusion, our bifurcation analysis allows one to state that a cou- ple composed of a secure and synergic individual and @ non-secure and non-synergic individual can have cyclic dynamics. 4. Discussion and Conclusion Dynamics of love between two persons has been investigated in this paper by means of a mini- mal model composed of two differential equations, one for each individual. Three mechanisms of love growth and decay have been taken into account: ‘The forgetting process, the pleasure of being loved and the reaction to partner’s appeal. This has been done by introducing two functions, called return and instinct functions, which differ in the cases of secure and non-secure individuals. The fact, here called synergism, that a woman [man] might re- act more strongly when she [he] is in love, has also been modeled. As a result, individuals are se- cure or non-secure, and synergic or non-synergic, for a total of four different classes. Thus, it has been shown that couples composed of non-synergic individuals as well as couples composed of secure2618 A. Gragnani et al. individuals cannot have cyclic dynamics. By con- trast, couples composed of a secure and synergic, individual and a non-secure and non-synergic indi- vidual car" ave cyclic dynamics. In other words, the « ‘ce of synergism and insecurity within the couple is the reason for tempestuous romantic relationships. This result has been proved through the analysis of the local and global bifurcations of the model. Fold, Hopf and homoclinic bifurcations have been detected, in accordance with the exis- tence of a Bogdanov-Takens codimension-2 bifur- cation point. The bifurcation diagrams can be easily inter- preted in terms of individual behavior and ap- peal. They show that, with the exception of a nar- row band in parameter space (regions 4 and 5 of Figs. 5-7), the system has a unique global attract ‘An equilibrium or a limit cycle. Both attractors are associated to positive feelings, i.e. to a satisfactory ‘quality of the romantic relationship. In general, the attractor is an equilibrium when the individuals are not too sensitive, ie. when their synergism and re- activeness to partner’s love and appeal are, as a whole, low. By contrast, high sensitivity (as well as high attractiveness) implies cyclic behavior, as one would intuitively expect. Our bifurcation diagrams also show that ageing has a stabilizing effect. Indeed, it is gener- ally believed that individual appeal, as well as syn- ergism and reactions to partner’s love and appeal, slowly deteriorate with ageing. Thus, couples with cyclic love dynamics (regions 3 and 4 of Figs. 5~7) can slowly vary during their life and finally become stationary by crossing the Hopf bifurcation curve ‘A or the homoclinic bifurcation curves p or q in Figs. 5-7. These two ways are qualitatively very dif- ferent and both are important. Indeed, they explain why cyclic romantic relationships characterized by relevant ups and downs tend to become more and more steady with the age of the partners. The Hopf bifurcation h says that in certain couples the ups and downs become less and less pronounced un- til they disappear completely, while the homoclinic bifurcation p and q says that cyclic romantic rela~ tionships can be very similar to a steady state inter- rupted from time to time by a burst of love. Getting closer and closer to the homoclinic simply implies that the time separating two subsequent bursts be- comes longer and longer, until the bursts definitely disappear when the homoclinic curve is crossed. As for any minimal model, the extensions one could propose are innumerable. In accordance with the most recent developments of attachment the- ory [Kobak & Hazan, 1991; Sharfe & Bartholomew, 1094), learning and adaptation processes could be taken into account allowing for some behavioral parameters to slowly vary in time. Suitable non- linearities could be introduced in order to develop theories for particular subclasses of non-secure in- dividuals (Griffin & Bartholomew, 1994]. Men and women could be distinguished by using different state equations [Hendrick & Hendrick, 1995]. The dimension of the model could also be enlarged in order to consider individuals with more complex personalities (Rinaldi, 1996b] or the dynamics of Jove in larger groups of individuals (e.g. families). Undoubtedly, all these problems deserve further attention. Acknowledgments This study has been financially supported by the Italian Ministry of Scientific Research and ‘Technology, under contract MURST 40% Teoria dei sistemi e del controllo. Part of the work has been carried out at the International Institute for Applied Systems Analysis, ITASA, Laxenburg, Austria. References Ainsworth, M., Blehar, M., Waters, E. & Wall, S. [1978] Patterns of Attachment (Erlbaum, Hillsdale NJ), Bartholomew, K. & Horowitz, L. M. [1991] “Attachment styles among young adults: A test of a four-category model,” J. Personality and Social Psychology 61, 226-244, Berman, A., Neumann M. & Stern, R. {1989} Non- negative Matrices in Dynamic Systems (Wiley, New York). Bowlby, J. [1969] Attachment and Loss: Attachment (Basic Books, New York) Vol. 1. Bowlby, J. (1973) Attachment and Loss: Separation (Basic Books, New York) Vol. 2. Bowlby, J. (1980} Attachment and Loss: Loss, Sadness, and Depression (Basic Books, New York) Vol. 3. Champneys, A. R. & Kuznetsov, Yu. A. [1994] “Numeri- cal detection and continuation of codimension-two ho- moclinie bifurcations,” Int, J. Bifurcation and Chaos 4, 785-822. Collins, N. L. & Read, S. J. [1990] “Adult attachment, working models and relationship quality in dating couples,” J. Personality and Social Psychology 68, 644-663. Feeney, J. A. & Noller, P. [1990] “Attachment. style as a predictor of adultic romantic relationships,” J. Personality and Social Psychology 58, 281-291.Feingold, A. [1990] “Gender differences in effects of phys- ical attractiveness on romantic attraction: A comper- ison across five research paradigms,” J. Personality and Social Psychology 59, 981-093. Puller, T. L. & Fincham, F. D. [1995] “Attachment style {in married couples: Relation to current marital func- tioning, stability over time, and method of assess- ment,” Personal Relationships 2, 17-34. Griffin, D. W. & Bartholomew, K. [1994] “Models of ‘the self and other. Fundamental dimensions underly- ing measures of adult attachment,” J. Personality and Social Psychology 67, 430445. Hendrick, 8. 8. & Hendrick, C. [1995] “Gender differences ‘and similarities in sex and love,” Personal Relation- ships 2, 55-65. Hazan, C. & Shaver, P. R. [1987] “Romantic love con- ceptualized as an attachment process,” J. Personality and Social Psychology 82, 511-524 Jones, F. J. [1995] The Structure of Petrarch’s Can- zoniere (Brewer, Cambridge). Khibnik, A. I, Kuznetsov, Yu. A., Levitin, V. V. & Nikolaev, E. V. [1993] “Continuation techniques and interactive software for bifurcation analysis of ODEs and iterated maps,” Physica D62, 360-371 Kobak, R. R. & Hazan, C. [1991] “Attachment in mar- riage: The effect of security and accuracy of work- ing models,” J. Personality and Social Psychology 60, 861-869. Luenberger, D. G. [1979] Introduction to Dynamic Sys- tems (Wiley, New York). Radzicki, M. J. [1993] “Dyadic processes, tempestuous ‘relationships, and system dynamics,” Syst. Dynam. Rev, 9, 79-94. Rinaldi, $. & Farina, L. [1995] Positive Linear Systems: Theory and Applications, In Italian. (Utet, Torino). Cyclic Dynamics in Romantic Relationships 2619 Rinaldi, S. 19968] “Love dynamics: The case of line ear couples,” Working Paper WP-96-68, Interna- tional Institute for Applied Systems Analysis, Laxen- bug, Austria (to appear in Applied Mathematics and Comprtaion). Rinaldi, S. [1996b] “Laura and Petrarch: An intrigu- ing case of cyclical love dynamics,” Working Paper ‘WP-96-67, International Institute for Applied Sys- tems Analysis, Laxenbug, Austria (to appear in SIAM J. Appl. Math). Rinaldi, S. & Gragnani, A. [1996] “Love dynamics between secure individuals: A modeling approach,” ‘Working Paper WP-96-69, International Institute for ‘Applied Systems Analysis, Laxenbug, Austria (to ap- pear es Dynamics, Psychology, and Life Sciences) Scharfe, E. & Bartholomew, K. [1994] “Reliability and stability of adult attachment patterns,” Personal Relationships 1, 23-43. Simpson, J. A., Gangestad, 8. W. & Lerma, M. {1990} “Perception of physical attractiveness: Mechanism in- volved in the maintenance of romantic relationships,” J. Personality and Social Psychology 58, 1192-1201. Sprecher, S., Quintin, S. & Hatfield, E. {1994} “Mate se- lection preferences: Gender differences examined in a national sample,” J. Personality and Social Psychol- ‘ogy 66, 1074-1080. Sternberg, R. J. & Baines, M. L. [1988} Psychology of Love (Yale U.P). Strogatz, S. H. [1988] “Love affairs and differential equa- tions,” Math. Mag. 61, 35. Strogatz, S. H. [1994] Nonlinear Dynamics and Chaos with Applications to Physics, Biology, Chemistry and Engineering (Addison-Wesley, Reading, MA).