The Blazhko effect is a long term, generally irregular modulation of the light curves that occurs in a sizeable number of RR Lyrae stars. The physical origin of the effect has been a puzzle ever since its discovery over a hundred years ago. We build here upon the recent observational and theoretical work of Szabo et al. on RRab stars who found with hydrodynamical simulations that the fundamental pulsation mode can get destabilized by a 9:2 resonant interaction with the 9th overtone. Alternating pulsation cycles arise, although these remain periodic, i.e. not modulated as in the observations. Here we use the amplitude equation formalism to study this nonlinear, resonant interaction between the two modes. We show that not only does the fundamental pulsation mode break up into a period two cycle through the nonlinear, resonant interaction with the overtone, but that the amplitudes are modulated, and that in a broad range of parameters the modulations are irregular as in the observations. This irregular behavior is in fact chaotic and arises from a strange attractor in the dynamics.

As shown by Mennickent, et al(2003), a subset of the blue variable stars in the Large Magellanic Cloud exhibit brightness variability of small amplitude in the period range 2.4 to 16 days as well as larger amplitude variability with periods of 140 to 600 days, with a remarkably tight relation between the long and the short periods. Our re-examination of these objects has led to the discovery of additional variability. The Fourier spectra of 11 of their 30 objects have 3 or 4 peaks above the noise level and a linear relation of the form f_a = 2(f_b - f_L) among three of the frequencies. An explanation of this relation requires an interplay between the binary motion and that of a third object. The two frequency relations together with the Fourier amplitude ratios pose a challenging modeling problem.

We review the current state of Cepheid modeling and discuss its dominant deficiency, namely the use of time dependent mixing length. Notwithstanding, Cepheid modeling has achieved some excellent successes, and we mention some of the most recent ones. Discrepancies between observations and modeling appear not so much in the gross properties of single mode Cepheids, but rather when more subtle nonlinear effects are important, such as in double mode or even triple mode pulsations. Finally we discuss what we consider the most important challenges for the next decade. These are, first, realistic multi dimensional modeling of convection in a pulsating environment, and second, the nonlinear modeling of the nonradial pulsations that have been observed, and, relatedly, of the Blazhko like phenomenon that has recently been observed in Cepheids.

A search for variable stars with ultra-low amplitudes (ULA), in the millimag range, has been made in the combined MACHO and OGLE data bases in the broad vicinity of the Cepheid instability strip in the HR diagram. A total of 25 singly periodic and 4 multiply periodic ULA objects has been uncovered. Our analysis does not allow us to distinguish between pulsational and ellipsoidal (binary) variability, nor between LMC and foreground objects. However, the objects are strongly clustered and appear to be associated with the pulsational instability strips of LMC Pop. I and II variables. When combined with the ULA variables of Buchler et al (2005) a total of 20 objects fall close to the classical Cepheid instability strip. However, they appear to fall on parallel period-magnitude relations that are shifted to slightly higher magnitude which would confer them a different evolutionary status. Low amplitude RV Tauri and Pop. II Cepheids have been uncovered that do not appear in the MACHO or OGLE catalogs. Interestingly, a set of binaries seem to lie on a PM relation that is essentially parallel to that of the RV Tauri/Pop. II Cepheids.

A number of microlensing dark-matter surveys have produced tens of millions of light curves of individual background stars. These data provide an unprecedented opportunity for systematic studies of whole classes of variable stars and their host galaxies. We aim to use the EROS-2 survey of the Magellanic Clouds to detect and study the population of beat Cepheids (BCs) in both Clouds. BCs pulsating simultaneously in the first overtone and fundamental modes (FO/F) or in the second and first overtone modes (SO/FO) are of particular interest. Using special software designed to search for periodic variables, we have scanned the EROS-2 data base for variables in the typical period range of Cepheids. Metallicities of FO/F objects were then calculated from linear nonadiabatic convective stellar models. We identify 74 FO/F BCs in the LMC and 41 in the SMC, and 173 and 129 SO/FO pulsators in the LMC and SMC, respectively; 185 of these stars are new discoveries. For nearly all the FO/F objects we determine minimum, mean, and maximum values of the metallicity. The EROS data have expanded the samples of known BCs in the LMC by 31%, in the SMC by 110%. The FO/F objects provide independent measures of metallicities in these galaxies. The mean value of metallicity is 0.0045 in the LMC and 0.0018 in the SMC.

It is well known that the mere location of a Beat Cepheid model in a Period Ratio \vs Period diagram (Petersen diagram) puts constraints on its metallicity Z. But these bounds are sensitive to the mixture of elements that are lumped into the parameter $Z$. In this short paper we update the previous results that were based on the Grevesse-Noels solar mixture to the recent, revised Asplund, Grevesse & Sauval (2005) (AGS) solar mixture. We also examine the effect of the envelope depth on the accuracy of the computed pulsation periods. We find that for low period Cepheids with high Z the customary approximation of envelope pulsation breaks down. It is necessary to compute stellar models that extend to the center and to include burning and composition inhomogeneities in the modeling. Fortunately, however, most Beat Cepheids that have been observed so far seem to avoid that regime.

We compute the phase lags between the radial velocity curves and the light curves $\Delta \Phi_1= \phi^{V_r}_1 - \phi^{mag}_1$ for classical Cepheid model sequences both in the linear and the nonlinear regimes. The nonlinear phase lags generally fall below the linear ones except for high period models where they lie above, and of course for low pulsation amplitudes where the two merge. The calculated phase lags show good agreement with the available observational data of normal amplitude Galactic Cepheids. The metallicity has but a moderate effect on the phase lag, while the mass-luminosity relation and the parameters of the turbulent convective model (time-dependent mixing length) mainly influence the modal selection and the period, which is then reflected in the period -- $\Delta \Phi_1$ diagram. We discuss the potential application of this observable as a discriminant for pulsation modes and as a test for ultra-low amplitudes (ULA) pulsation.

The mere location of a Beat Cepheid model in a Period Ratio vs. Period diagram (Petersen diagram) puts very tight constraints on its metallicity Z. The Beat Cepheid Peterson diagrams are revisited with linear nonadiabatic turbulent convective models, and their accuracy as a probe for stellar metallicity is evaluated. They are shown to be largely independent of the helium content Y, and they are also only weakly dependent on the mass-luminosity relation that is used in their construction. However, they are found to show sensitivity to the relative abundances of the elements that are lumped into the metallicity parameter Z. Rotation is estimated to have but a small effect on the 'pulsation metallicities'. A composite Petersen diagram is presented that allows one to read off upper and lower limits on the metallicity Z from the measured period P0 and period ratio P1/P0.

Our analysis of the Deep CFHT M33 variability survey database has uncovered 5 Beat Cepheids (BCs) that are pulsating in the fundamental and first overtone modes. With {\it only} the help of stellar pulsation theory and of mass--luminosity (M-L) relations, derived from evolutionary tracks, we can accurately determine the metallicities Z of these stars. The [O/H] metallicity gradient of -0.16 dex/kpc that is inferred from the M33 galacto-centric distances of these Cepheids and from their 'pulsation' metallicities is in excellent agreement with the standard spectroscopic metallicity gradients that are determined from H II regions, early B supergiant stars and planetary nebulae. Beat Cepheids can thus provide an additional, independent probe of galactic metallicity distributions.

The MACHO variables of LMC Field 77 that lie in the vicinity of the Cepheid instability strip are reexamined. Among the 144 variables that we identify as Cepheids we find 14 that have Fourier amplitudes <0.05 mag in the MACHO red band, of which 7 have an amplitude <0.006 mag : we dub the latter group of stars ultra-low amplitude (ULA) Cepheids. The variability of these objects is verified by a comparison of the MACHO red with the MACHO blue lightcurves and with those of the corresponding OGLE LMC stars. The occurrence of ULA Cepheids is in agreement with theory. We have also discovered 2 low amplitude variables whose periods are about a factor of 5--6 smaller than those of F Cepheids of equal apparent magnitude. We suggest that these objects are Cepheids undergoing pulsations in a surface mode and that they belong to a novel class of Strange Cepheids (or Surface Mode Cepheids) whose existence was predicted by Buchler et al. (1997).

We describe a methodology that allows us to follow the pulsational behavior of an RR Lyrae model consistently and automatically along its evolutionary track throughout the whole instability strip. It is based on the powerful amplitude equation formalism, and resorts to a judicious combination of numerical hydrodynamical simulations, the analytical signal time-series analysis, and amplitude equations. A large-scale survey of the nonlinear pulsations in RR Lyr instability strip is then presented, and the mode selection mechanism is delineated throughout the relevant regions of parameter space. We obtain and examine two regions with hysteresis, where the pulsational state depends on the direction of the evolutionary tracks, namely a region with either fundamental (RRab) or first overtone (RRc) pulsations and a region with either fundamental (RRab) or double-mode (RRd) pulsations. The regions where stable double-mode (DM, or RRd) pulsations occur are very narrow and hard to find in astrophysical parameter (L, M, T_eff, X, Z) space with hydrodynamic simulations, but our systematic and efficient methodology allows us to investigate them with unprecedented detail. It is shown that by simultaneously considering the effects of mode selection and of horizontal branch evolution we can naturally solve one of the extant puzzles involving the topologies of the theoretical and observed instability strips, namely the slope of the fundamental blue edge. The importance of the interplay between mode selection and stellar evolutionary effects is also demonstrated for the properties of double-mode RR Lyr. Finally, the Petersen diagram of double-mode RR Lyr models is discussed for the first time.

An analysis of the photometric observations of the light curves of the five large amplitude, irregularly pulsating stars R UMi, RS Cyg, V CVn, UX Dra and SX Her is presented. First, multi-periodicity is eliminated for these pulsations, i.e. they are not caused by the excitation of a small number of pulsation modes with constant amplitudes. Next, on the basis of energetics we also eliminate stochasticity as a cause, leaving low dimensional chaos as the only alternative. We then use a global flow reconstruction technique in an attempt to extract quantitative information from the light curves, and to uncover common physical features in this class of irregular variable stars that straddle the RV Tau to the Mira variables. Evidence is presented that the pulsational behavior of R UMi, RS Cyg, V CVn and UX Dra takes place in a 4-dimensional dynamical phase space, suggesting that two vibrational modes are involved in the pulsation. A linear stability analysis of the fixed points of the maps further indicates the existence of a two-mode resonance, similar to the one we had uncovered earlier in R Sct: The irregular pulsations are the result of a continual energy exchange between two strongly nonadiabatic modes, a lower frequency pulsation mode and an overtone that are in a close 2:1 resonance. The evidence is particularly convincing for R UMi, RS Cyg and V CVn, but much weaker for UX Dra. In contrast, the pulsations of SX Her appear to be more complex and may require a 6D space.

The pulsational properties of the Cepheid models along the evolutionary tracks from the Padova group (Girardi et al.), as calculated with our turbulent convective pulsation code, are in good agreement with the resonance constraints imposed by the observational OGLE-2 data of the Small and Large Magellanic Clouds. Our study suggests that the P_4/P_1=1/2 resonance for the overtone Cepheids occurs for periods clustering around 4.2 d, in disagreement with the suggestion of Antonello & Poretti based on the observations of light curves, but in agreement with Kienzle et al. and Feuchtinger et al. For the fundamental Cepheids the lowest order Fourier decomposition coefficients from the light curves, \viz R_21 and phi_21 can be used to locate the resonance region, but not so for the first overtone Cepheids. Here, the radial velocity curves can be used to locate the overtone resonance region, or in their absence, one needs to resort to numerical hydrodynamic modelling.

We demonstrate how, with a purely empirical analysis of the irregular lightcurve data, one can extract a great deal of information about the stellar pulsation mechanism. An application to R Sct thus shows that the irregular lightcurve is the result of the nonlinear interaction of two highly nonadiabatic pulsation modes, namely a linearly unstable, low frequency mode, and the second mode that, although linearly stable, gets entrained through a 2:1 resonance. In the parlance of nonlinear dynamics the pulsation is the result of a 4 dimensional chaotic dynamics.

Near the edges of the instability strip the rate of stellar evolution is larger than the growth-rate of the pulsation amplitude, and the same holds whenever the star is engaged in pulsational mode switching. Stellar evolution therefore controls both the onset of pulsation at the edges of the instability strip and of mode switching inside it. Two types of switchings (bifurcations) occur. In a soft bifurcation the switching time scale is the inverse harmonic mean of the pulsational modal growth-rate and of the stellar evolution rate. In a hard bifurcation the switching times can be substantially longer than the thermal time scale which is typically of the order of a hundred periods for Cepheids and RR Lyrae. We discuss some of the observational consequences, in particular the paucity of low amplitude pulsators at the edges of the instability strip.

The numerical hydrodynamic modelling of beat Cepheid behavior has been a long standing quest in which purely radiative models had failed consistently. We find that beat pulsations occur naturally when {\it turbulent convection} is accounted for in our hydrodynamics code. The development of a relaxation code and of a Floquet stability analysis greatly facilitates the search for and the analysis of beat Cepheid models. The conditions for the occurrence of beat behavior can be understood easily and at a fundamental level with the help of amplitude equations.

La plupart des etoiles variables qui pulsent avec grande amplitude, telles les cepheides, ont un comportement tres regulier et periodique. Mais juste a cote d'elles dans le diagramme Hertzsprung-Russell, se trouve un groupe d'etoiles variables dotees de courbes de lumiere tres irregulieres. A l'aide d'une technique de reconstruction de flot appliquee aux donnees d'observation astronomiques de plusieurs de ces etoiles, on montre que la dynamique sous-jacente est chaotique et de faible dimension, ce qui peut surprendre a cause de la violence de ces pulsations. En plus a l'aide d'une linearisation du flot on deduit que le mecanisme physique de la pulsation consiste en l'interaction entre deux modes vibratoires (vraisemblablement radiaux), l'un lineairement instable, de frequence f_0, et l'autre stable, mais de frequence \~2f_0 (soit un scenario a la Shilnikoff generalise).

The light curves (time series of the radiated energy) of most large amplitude, pulsating stars such as the well known Cepheid stars are regular. However, a smaller group of variable stars that are located next to them in the Hertzsprung- Russell diagram undergoes irregular light variations and exhibits irregular radial velocities as well. The mechanism behind this irregular behavior was a long standing mystery. A flow reconstruction technique based on the observed lightcurves of six separate stars shows that their underlying dynamics is chaotic and low dimensional (d=4). Furthermore, we present evidence that the physical mechanism behind the behavior is the nonlinear interaction of just two pulsation eigenmodes. In a generalized Shil'nikov scenario, the pulsation energy alternates continuously, but irregularly between a lower frequency mode that is linearly unstable and thus growing, and a stable overtone that gets entrained through a low order resonance (2:1), but that wants to decay. The flow reconstruction from the stellar light curve thus yields interesting physical insight into the pulsation mechanism.

The OGLE data base is used in conjunction with Kurucz atmosphere models to generate sets of period, effective temperature and luminosity for fundamental and overtone Magellanic Cloud Cepheids. The Florida pulsation code (with linear turbulent convection) is then used to compute masses for these stars, assuming an average composition of ($X$=0.716, $Z$=0.010) for the LMC and of ($X$=0.726, $Z$=0.004) for the SMC. The average $M$--$L$ relation for the fundamental Cepheids matches closely that for the first overtone Cepheids for each Magellanic Cloud. Neither the SMC nor the LMC average $\Log M$--$\Log L$ relations are straight, but have a noticeable curvature. In view of the uncertainties in distance and reddening we have adopted three different choices for these quantities. The results based on the 'long' distance scale to the clouds give a better agreement between theory and and observations than the 'short' one. All the current evolutionary tracks predict systematically larger masses for given luminosities than our observationally derived ones, especially at the high end. Moreover, our study confirms that the evolutionary tracks of the low mass stars in SMC are not in agreement with the observations as they do not extend sufficiently blueward and do not penetrate deep enough into the instability strip, or not at all. The inference of masses directly from the observational database yields a novel and strong constraint on evolutionary calculations.

Strange modes can occur in radiative classical Cepheids and RR Lyrae models. These are vibrational modes that are trapped near the surface as a result of a 'potential barrier' caused by the sharp hydrogen partial ionization region. Typically the modal number of the strange mode falls between the 7th and 12th overtone, depending on the astrophysical parameters of the equilibrium stellar models (L, M, \Teff, X, Z). Interestingly these modes can be linearly unstable outside the usual instability strip, in which case they should be observable as new kinds of variable stars, 'strange Cepheids' or 'strange RR Lyrae' stars. The present paper reexamines the linear stability properties of the strange modes by taking into account the effects of an isothermal atmosphere, and of turbulent convection. It is found that the linear vibrational instability of the strange modes is resistant to both of these effects. Nonlinear hydrodynamic calculations indicate that the pulsation amplitude of these modes is likely to saturate at the millimagnitude level. These modes should therefore be detectable albeit not without effort.

A hydrodynamical survey of the pulsational properties of first overtone Galactic Cepheids is presented. The goal of this study is to reproduce their observed light- and radial velocity curves. The comparison between the models and the observations is made in a quantitative manner on the level of the Fourier coefficients. Purely radiative models fail to reproduce the observed features, but convective models give good agreement. It is found that the sharp features in the Fourier coefficients are indeed caused by the P1/P4 = 2 resonance, despite the very large damping of the 4th overtone. For the adopted mass-luminosity relation the resonance center lies near a period of 4.2d +/- 0.2 as indicated by the observed radial velocity data, rather than near 3.2d as the light-curves suggest.

The luminosities, effective temperatures and metallicities that are derived empirically by Kovacs and Jurcsik from the light curves of a large number of globular cluster and field RRab and RRc stars are compared to theoretical RR Lyrae models. The strong luminosity dependence of the empirical blue and red edges (Log L vs Log Teff diagram) is in disagreement with that of both radiative and convective models. A reexamination of the theoretical uncertainties in the modelling leads us to conclude that the disagreement is irreconcilable.

The status of the hydrodynamical modelling of nonlinear multi-mode stellar pulsations is discussed. The hydrodynamical modelling of steady double-mode (DM) pulsations has been a long-standing quest that is finally being concluded. Recent progress has been made thanks to the introduction of turbulent convection in the numerical hydrodynamical codes which provide detailed results for individual models. An overview of the modal selection problem in the HR diagram can be obtained in the form of bifurcation diagrams with the help of simple nonresonant amplitude equations that capture the DM phenomenon.

The Fourier spectral techniques that are common in Astronomy for analyzing periodic or multi-periodic light-curves lose their usefulness when they are applied to unsteady light-curves. We review some of the novel techniques that have been developed for analyzing irregular stellar light or radial velocity variations, and we describe what useful physical and astronomical information can be gained from their use.

We review the numerical modelling of the nonlinear pulsations of classical variable stars with hydrocodes that include the effects of turbulent convection. Despite their simplicity these turbulent convective recipes appear to remove many of the difficulties that radiative codes faced. In particular, the numerical modelling of double mode pulsations has become possible.

We give a status report of convective Cepheid and RR Lyrae model pulsations. Some striking successes can be reported, despite the use of a rather simple treatment of turbulent convection with a 1D time-dependent diffusion equation for the turbulent energy. It is now possible to obtain stable double-mode (beat) pulsations in both Cepheid and RR Lyrae models with astrophysical parameters, i.e. periods and amplitude ratios, that are in agreement with observations. The turbulent convective models, however, have difficulties giving global agreement with the observations. In particular, the Magellanic Cloud Cepheids, that have been observed in connection with the microlensing projects have imposed novel observational constraints because of the low metallicity of the MCs.

The mechanism for efficient and coherent angular momentum transport remains one of the unsolved puzzles in astrophysics despite the enormous efforts that have been made. We suggest that important new insight could be gained in this problem through an experimental and theoretical study of a laboratory device (Ranque-Hilsch tube) that displays a similar enhanced angular momentum transfer which cannot be explained by a simple turbulent model. There is already good experimental evidence to suggest that the cause of this enhancement is the formation of aligned vortices that swirl around the symmetry axes very much like virtual paddle blades.

We review recent results of stellar pulsation modelling that show that even very simple one-dimensional models for time dependent turbulent energy diffusion and convection provide a substantial improvement over purely radiative models.

The numerical hydrodynamic modelling of beat Cepheid behavior has been a longstanding quest in which purely radiative models have failed miserably. We find that beat pulsations occur naturally when turbulent convection is accounted for in our hydrodynamics codes. The development of a relaxation code and of a Floquet stability analysis greatly facilitates the search for and analysis of beat Cepheid models. The conditions for the occurrence of beat behavior can be understood easily and at a fundamental level with the help of amplitude equations. Here a discriminant D arises whose sign decides whether single mode or double mode pulsations can occur in a model, and this D depends only on the values of the nonlinear coupling coefficients between the fundamental and the first overtone modes. For radiative models D is always found to be negative, but with sufficiently strong turbulent convection its sign reverses.

A one-dimensional turbulent convection model in the form of a time-dependent diffusion equation for the turbulent energy is incorporated into our numerical pulsation code. The effect of turbulent convection on the structural rearrangement of the static equilibrium star is taken into account, and the linear eigenvalues (periods and growth-rates) for the complete turbulent convective nonadiabatic pulsations of the equilibrium models are calculated. The linearized code allows us to perform efficiently a systematic survey of the stability of Cepheid models for a wide range of astrophysical parameters (L, M, Teff) and the turbulent convective parameters. A sensitivity analysis of the properties of several sequences of Galactic Cepheids is performed with the goal of using observational constraints, such as the shape and location of the fundamental and first overtone instability strips, to calibrate the turbulent convective input parameters. The locations of the red and blue edges of the fundamental and first overtone instability strips and the overall strip widths are largely determined by the parameters that determine the mixing length, and strengths of the eddy viscosity and of the convective flux. The remaining parameters can be used for fine-tuning agreement with observations.

Strange modes have been found in the radial spectrum of many luminous stars, most recently in Cepheids and RR Lyrae. We show that there is nothing strange about these modes -- they must exist even in the adiabatic limit. With a change of variables and without approximation the adiabatic linear pulsation equation is reduced to a Schroedinger equation in which the radial coordinate is the local sound-traversal time. In this formulation, the narrow hydrogen partial ionization region is seen to act as a potential barrier, separating the star into two regions. Resonances between the two regions result in the strange modes, for which the ratio of interior to exterior amplitude is at a minimum. The relative location of the barrier changes with the stellar parameters, and this gives rise to avoided level-crossings along a sequence of models. 2 The appearance of strange modes and the associated level crossings are exhibited with an analytic toy-model with the potential barrier approximated by a delta function. This toy-model is readily extensible to nonadiabatic modes. Hydrodynamical calculations find that pure strange mode limit-cycles have extremely small photospheric velocities and luminosity variations in the milli-magnitude range. They are therefore expected to be difficult to observe.

The light curves of the two large amplitude variable stars, R Scuti and AC Herculis, display irregular pulsation cycles with 'periods' of 75 and 35 days, respectively. We review the evidence that the observed time-series are generated by low dimensional chaotic dynamics. In particular, a global flow reconstruction technique indicates that the minimum embedding dimensions are 4 and 3 for R Sct and AC Her, respectively, whereas the fractal dimensions are inferred to be approximately 3.1 and 2.05 - 2.45, respectively. It thus appears that the dimensions of the dynamics themselves are also 4 and 3.

The Galactic Cepheid period histogram has a strong dip between 8 and 10 days that has defied an explanation based on evolutionary and linear pulsation studies. We show here that this deficiency is caused by the instability of the nonlinear fundamental pulsation cycle in this period range. The strong metallicity dependence of this instability is consistent with the absence of a corresponding minimum in the Magellanic Cloud data. Our results also suggest that the Galactic Cepheids must have a large spread in metallicity.

We review some of the recent advances in nonlinear pulsation theory, but also insist on some of the major extant shortcomings.

The Fourier spectral analysis of this object suggests a nonlinear pulsation of period approximately 14d with a 3:2 frequency lock between two vibrational modes. We discuss an extensive search that we have performed in the L, M, Teff space for models of period 14 - 16 days that exhibit this particular resonance. We find models satisfying the resonance constraints both in the Cepheid and in the post AGB regimes. The nonlinear pulsations of several candidates are presented. The nature of the star remains a mystery as none of the models satisfies all the observational constraints.

We describe an implicit 1--D adaptive mesh hydrodynamics code that is specially tailored for radial stellar pulsations. In the Lagrangean limit the code reduces to the well tested Fraley scheme. The code has the useful feature that unwanted, long lasting transients can be avoided by smoothly switching on the adaptive mesh features starting from the Lagrangean code. Thus, a limit cycle pulsation that can readily be computed with the relaxation method of Stellingwerf will converge in a few tens of pulsation cycles when put into the adaptive mesh code. The code has been checked with two shock problems, viz Noh and Sedov, for which analytical solutions are known, and it has been found to be both accurate and stable. Superior results were obtained through the solution of the total energy (gravitational + kinetic + internal) equation rather than that of the internal energy only.

We briefly review the evolutionary status of the classical Cepheid variables, their structure, and the properties of the linear modes. Then we discuss the current status of the nonlinear hydrodynamical modelling, including modern adaptive mesh techniques, and the sensitivity to opacity. We stress the importance of resonances among the normal modes of pulsation because they can be used to impose severe constraints on the models and on the input physics. We show that recent hydrodynamics computations provide an explanation for the hitherto unexplained dip in the histogram of Galactic Cepheids around 8--10 days. The EROS and MACHO observations of classical Cepheids in the Large and Small Magellanic Clouds have opened up a new exciting perspective on these variable stars. This new observational basis is very important, not only because of the sheer number of new Cepheids, but especially because the composition of the MCs is known to differ from our Galaxy's, thus extending the range of astrophysical parameters. These new data have been found not to be fully compatible with our current Cepheid modelling and are forcing a review of our assumptions and input physics.

It is first shown that the observational light curve data of R Scuti, a star of the RV Tau type, is not multi-periodic, and that it cannot have been generated by a linear stochastic (AR) process. By default, the signal must be a manifestation of deterministic chaos. We use a novel nonlinear time-series analysis, the global flow reconstruction technique, to probe the properties of the irregular pulsation cycles. We show in particular that the chaotic dynamics of this star's complicated lightcurve is captured by a simple 4D polynomial map or flow (4 first order ODEs). Importantly also, the method allows us to quantify an irregular signal which has the potential benefit for extracting novel stellar constraints from an irregular light-curve. Finally, from the low dimensionality 4 of the flow we can infer a simple physical picture of the pulsations, and arguments are presented that the pulsations of R Sct are the result of the nonlinear interaction of two vibrational normal modes of the star.

We introduce the reader to the "global flow reconstruction method". The purpose of the method is to see if a given temporal sequence has been generated by a low dimensional dynamics, and to determine that dimension and other properties of the dynamics. The method is shown to work very well on the well known Roessler system. An application of the method to the observational data of the light curve of the irregular variable star R Scuti is presented. This light curve is shown to have been generated by a 4 dimensional dynamics. From the functional form of the flow we find that the Lyapunov dimension is close to 3.1.

The AAVSO lightcurve data of the irregularly pulsating star AC Herculis of the RV Tau class are analysed. The lightcurve is shown to be incompatible with a periodic, or even multiperiodic pulsation, even if allowance is made for evolution. Instead the best explanation is that the irregularly alternating cycles are a manifestation of low dimensional chaos. The lightcurve is found to be generated by a 3 or 4 dimensional dynamics -- 3 or 4 first order ODEs. The (Lyapunov) fractal dimension of the underlying dynamic attractor is computed to be d_L ~ 2.2, smaller than the value of d_L ~ 3.1 found for R Sct.