Referee 1 1. a) Comparison with the the TAPS and KAOS data has been added, see new Fig. 10. and its description in text has been added (table with integral cross sections can be added, too but data agree within arrors and difference is below 10%). b) Adding this figure, the comparison of the pi multiplicities per participants for our and TAPS/KaoS (recalculated) data is not needed any more, and has been removed from Table 4. c) TAPS two slopes Fig.3 in ref.21 Averbeck et al PR67 024903 (2003) really shows a weak deviation from the one-slope fit at low transverse masses when they fit the distribution in region above 400 MeV/c. As seen from Fig. 10 (inset with ratio), the effect is much stronger for charged mesons. We fitted the pi0 distribution in full range, with result of chi2/(deg. of. freedom) to be 68/29 for 1 slope and 38/27 for 2 slopes, respectively. The improvement by 2 slope fit is only slight and not very significant. d) we use now reference to the original TAPS paper. 2. Extrapolation of the rapidity distribution outside the HADES acceptance (section 4.2) has been done using PLUTO (according to the recommendation). Systematic error were estimated by varying the input parameters into PLUTO within their experimental errors. 3. To estimate quality of the LVL1 trigger selection implemented in UrQMD we checked both the charged hit (in META) and charged track multiplicities. The agreement was in the same order of several percents in the mean, and slightly different shape of distribution. The reason is that our analog multiplicity trigger is not perfect, and cannot be simulated precisely. Moreover number of secondary particles produced in interactions of primary particles with material is not simulated perfectly. For the presentation we selected the charged track multiplicity, as it reflects the centrality better than the charged hits only. Systematic error estimate of the centrality selection has been added in Section 3.5. 4. anisotropy a) "The comparison of the results reflected about y0 is an important check, but this check suggests that the systematic errors are larger than estimated." Chi2 per number of degrees of freedom is close to one, so it shows that errors are realistic. b) Figures 11 - 13 were made larger. c) Fig.12 (former Fig.11) UrQMD results have been added. Dashed line has been explained in the figure caption. d) "Since the errors presumably are dominated by systematic errors, it would be useful to further divide the data sample based on centrality, i.e. track multiplicity, to investigate the dependence of the anisotropy on centrality." We did this investigation, dividing events into several bins in track multiplicity. We did not observe any significant change in the dependence of the asymetry on momnetum. The reason is that in C+C system the centrality selection is very weak; even if we select events with high track multiplicity, we still have in our sample events with very broad range of impact parameters, as it is seen in Fig.6. e) Fig.13 (former Fig.12) "First it appears that data points are missing for the case of 1AGeV pi+ at 550 MeV/c and for 2AGeV pi+ at 650 MeV/c - comparing to pi-. Is there a reason?" The angular distribution for pi+ in these mom. bins is asymmetric, showing much more pi+ in the "TOFINO" region - it shows that pi+ - p separation is not good enough. Therefore the fit (by A0*(1.+A2cos(theta_cms)) is bad, and the A2 result is difefrent (higher) than for pi-. Therefore we removed these points from the figure. f) Errors of UrQMD points in Fig.13 are small and not visible (a sentence also added into Fig. caption). 5. Summary ????? 6. Systematic errors from the efficiency uncertainties: "It was stated that systematic errors were estimated based on comparisons of the results from different sectors of the HADES spectrometer. Are the measurements from all sectors entirely independent? Are there common potential systematic errors - like the absolute field measurements, tracking chamber geometry, or PID yield extraction procedures, that are not reflected in sector comparisons." The errors based on comparison of different sectors are rather large, about 5%. There are other sources of errors, as stated by the referee. They are estimated to be much smaller from the simulations of the experiment, including full analysis chain, and also from the analysis of the proton-proton elastic scattering from other HADES experiment. 7. "The number of UrQMD events simulated should be stated. Are the statistical errors on the UrQMD results negligible? What are the systematic errors resulting from the centrality selection?" The number of simulated UrQMD events has been added in section 3.1. Statistical errors are negligible. Systematic errors of integral yields resulting from the centrality selection are estimated as in section 3.5 as 7%. The distribution shapes change very weakly with the change of centrality within these estimated limits, and are neglected. Especially we checked that for the polar anisotropy study, and did not observe any effect. For example, in case of UrQMD at 2A GeV, the value of the A2 asymmetry coefficient changes only from 1.12 to 1.16 coming from the LVL1 data to min. bias (fitted within HADES acceptance). Referee 2 1. Introduction has been modified. a) The first sentence has been rephrased and more references have been added. b) The UrQMD model is indeed a physical model which describes successfuly main features of the heavy-ion reactions. The description of transport models has been added to introduction. 2. On the other hand, the PLUTO generator is just a simple Monte-Carlo generator, which produces particles distributed according to the initial parameters of the thermal source. The observables experimentaly deduced in this paper (inverse slopes, anisotropies) are used as initial parameters. In this paper PLUTO is used only to produce the pion rapidity distribution, which allow to extrapolate out measured data outside the HADES acceptance. No physics conclusions are drawn from the comparison of PLUTO and measured spectra. PLUTO is now introduced in the section 3.1 (Simulation). 3. Table 2 (m_t fit results) All fits have been done taking into account realistic data points errors. The statistical errors are negligible. The systematic errors coming mainly from the efficiency uncertainties were estimated from differences of distributions from 6 independent sectors (both for measured and simulated data). The resulted chi2 normalized to number of degrees of freedom close to one, so the obtained errors can be trusted. Much larger errors of the inverse slope parameters from the two-slopes fit with respect to one-slope fit are caused by large correlations between fit parameter in the first case. In case of 1 GeV data we put the result of fit by two slopes to the table only to be consistent with the 2 GeV case. It only showed that the 2nd slope is not needed, the fit by one slope is statisticaly very good. We have removed the two-slopes fit from the table. The "pion species" have been replaced by "pi mesons" 4. Summary We believe that there is new information on physics coming from comparison of our data with the UrQMD calculations. As we state, the overall agreement of data and UrQMD is rather good, for yields, anisotropies as well as rapidities also within still limited measured range, On the other hand we clearly observe differences of data and UrQMD, namely a) Two-slopes feature od m_t at 2GeV for pi+ and pi-, not observed in UrQMD. This differs also from the same distribution of pi0 (TAPS). b) All transport models predict excess of pi0 over pi+/pi- yield, which we do not observe experimentaly. Unfortunately rather large errors do not allow to make a strict conclusion.