Specific comments/questions:

Abstract

" A|| for the deuteron has been extracted from these data as a function of W and Q2 with unprecedented precision" should be replaced by " A|| for the deuteron has been extracted in the valence region from these data as a function of W and Q2 with unprecedented precision"
Reply: Done.

Data analysis
- page 6, table 3: are the quantities labelled "geoAve" or "ariAve" used at all in the analysis? My understanding is that the average Q2 is calculated for each bin of each data set, averaging the Q2 of the selected event, i.e. weighting with the actual distribution. Is this correct?
Reply: This Q2 binning is used for plots. The average Q2 for each data point is calculated from data and data points within the same Q2 are averaged together. The arithmetic Q2 values, ariAve, are used only for the plots. The geoAve values are not used anywhere. 

- page 7, 2nd. par: Eq.8 repeats the Eq. 2. I suggest just to refer to Eq.2.
Reply: Done.

- sec 1.3:
-- I think a legend, describing the various variables, would be useful for Equations (12) and (13), or at least refer to the corresponding page in the proton note.
Reply: I added a table (Table 4) for the description of various parameters in these equations.

-- a factor of 6/21, corresponding to the number of polarized nucleons over the total, is used in the calculation of the dilution factor; is therefore Sigma_N defined as the average of the proton and neutron cross sections?
Reply: For the cross-section of 15N, models from Peter Bosted, which was partly generated from the 15N/12C target runs in EG1b were used. For detailed explanations, please refer to the paper by P. Bosted and R. Fersch {\it et al}, "Ratios of $^{15}$N/$^{12}$C and $^4$He/$^{12}$C inclusive electro-production cross-sections in the nucleon resonance region", \prc{78} 015202 (2008) [arXiv:nucl-ex/0712.2438v3]. 

- page 10, fig. 1: the value of L shows a non negligible dependence on W: was this taken into account in the evaluation of systematic errors? Is the size of the systematic error making this dependence (as well as the one visible in the following plots) negligible?
Reply: Varying the W-cut used for L was considered in the systematic errors and is part of the cited error for the dilution factor. At high W, the L is most certainly inaccurate due to low-momentum effects; these values are cut, and the upper bound was varied considerably to measure this effect. The error for average L is about 1.5% of A_||.


- sec. 1.4.2:

-- page 20, after Eq.18: "The two methods yielded very similar results" what is the momentum dependence in %?
***A***:

-- page 22, and also in the following, you say that you determine a scaling factor for the carbon using the phi distribution of quasi elastic events. Why not use the ratio of FC counts times ratio of target lengths for the two data sets (this is how we do in eg1-dvcs)?
***A***:

-- how were the kinematic cuts chosen?
***A***: To select the quasi-elastic events, we first need to find the scattered electron with these kinematic cuts. They are explained in detail in the proton analysis note, therefore we removed it from this note to make it more compact.

-- what is the meaning of "flag" in table 10?
Reply: Status Flag: Each identified particle in the DST carries a status flag. If the particle status flag is in the [0,5] range, the reconstruction is time-based and the event is acceptable. Particles with status flag $>$ 5 are reconstructed from hit-based tracking only and should be eliminated. In addition, if a particle is detected in all 3 super-layers of the DC and its trajectory is reconstructed accurately, 10 is added to the status flag variable, which carries some of the time based tracks into a range of [10,15]. Therefore, the listed cut was applied for the status flag. 

-- in the discussion of the inclusive method, it is stated (second to last paragraph of page 20) that two methods were used to account for the D-wave: what is this second method? Can you quantify the difference?
***A***:

-- in page 21, referring to the inclusive method, it is stated that the W cuts were varied to find a stable region: can you show plots?
***A***:

-- in the exclusive method, how much does the result depend on the phi region used for the BG subtraction?
***A***:

-- in table 12, it is mentioned that events with one neutral were also selected: what is the fraction of those events?
***A***:

-- the values in table 14 do no match the values reported on the plots in the previous pages: why?
***A***:

-- in the caption of table 14 it is said that for the 1.6 and 1.7 outbending data the inclusive method results were used, with a 10% error. But by looking at Fig. 12, bottom plot, we see that for E=1.723 outbending there is a big difference between the two methods (~0.25 for excl and ~0.17 for incl), that goes way beyond a 10% factor. What is the reason for choosing the inclusive method? In the figure it seems to have worse statistical precision and shows some Q2 dependence. Does this data set (and the others were the results of the two methods differ considerably) have a bigger systematic uncertainty on PbPt due to this discrepancy?
***A***:


- Table 15 is not referred to in the text, if I didn't miss it somehow. What is the "excl 2" method? I didn't find its description in the text.
***A***:

- The section 1.5 on polarized background relies on the CLAS-Note by M. Mayer, which was not reviewed before. The conclusion on the proton background being unpolarized needs more support. It is absolutely crucial for future studies at 12 GeV with ND3 target, to understand the sources of possible contaminations of ND3 with polarized and unpolarized protons, and any existing data set is important.
***A***:

- section 1.5: can you quantify the final values of the parameters discussed in this section?
***A***:
Values used for the analysis:
  C1 =  1.030; 
  C2 =  0.039;           
Variation for the sys error for polarized nitrogen corrections for ND3
  C1 = 1.020; 
  C2 = 0.027; 
Correction applied to A_||.
  Tapar_qw_el[q][w] = C1*(Tapar_qw_el[q][w] - C2*AparP_unrad[q][w]);
  TaparErr_qw_el[q][w] = C1*(TaparErr_qw_el[q][w]  - C2*AparErrP_unrad[q][w]);                



- The statement on page 34 "Here, the lower estimate is the standard assumption made for similar targets in past experiments [120] (assuming, e.g., that approximately 4.5% of the ND3 molecules are actually ND2H1)" refers to K.Abe et al, where the ND2H1 was not mentioned.
***A***:

- in the middle of page 35, you mention the "second term in either form of Cd", what are you referring to? In the form of Eq. (28) I don't see second terms.
***A***: This meant to refer to the term Dn/(1-1.5wD) but as you said, the expression is misleading, so I changed this in the document as well.

- in page 35, last paragraph: "its uncertainty is included in the overall uncertainty of Dn." What is the fraction of EMC part in overall error on Dn? SLAC paper uses 1/3 factor while we use 1/2.5 in Eq 31 (factor P14N/Pd is estimated via the equal spin temperature hypothesis to be between 0.4 and 0.5). From the discussion below (also in page 36), it is not clear whether the part with Dn was accounted for or not. Where is the C1^d=1.03 value coming from? Is the quoted 5% hydrogen contamination considered polarized in the extraction of PbPT?
***A***: Dn is considered in these calculations. C1d = 1.03 is the value using the average values of each variable in the equation. The error is estimated using the boundary values. 
More detailed explanation to be written here later with the calculations themselves..

- page 37, end of first paragraph, I think you missed brackets in the definition of the range of Cd2.
Reply: Corrected.

- bottom of page 37, how do you get the value 62%?
***A***:

- section 1.6: What are the results for the radiative corrections? What are the obtained values of f_RC and A_RC? A table or some plots, or even a mention of the size of the corrections would be nice.
***A***: There are data tables, which are quite long. I can create a plot of f_RC and A_RC as a function of W for each Q2 bin (and for each data set). The crude impression from the data tables suggest that f_RC ranges from 0.2 to 0.9 as Q2 increases. A_RC ranges from 0.01 to 0.1 as W and Q2 increases. I observe higher A_RC in the elastic threshold region on the order of 0.2 - 0.5 depending on the data set.

- page 41: referring to HERMES fit (which is for F_2) the F_1 are plotted in Fig. 15. How HERMES fit was used? How the extracted parametrization for A1 discussed in page 43 compares to known PDF fits?
***A***:


- section 1.8:
-- page 46: definitions of several variables (\eta,\eps) appear first here, while they were used long before.
Reply: I put the definition of several variables to the introduction section as well.

-- no T-test is done when the asymmetries from opposite half-wave plate runs are combined? Why?
Reply: We are not combining asymmetries from opposite hwp, we combine the counts. Please see the first paragraph of section I.8.1, where I tried to explain this in a bit more detailed way.

-- in the evaluation of the t-Test, was the error on the asymmetries statistical only?
Reply: Yes, only the statistical errors were used for the t-Tests.

-- in several cases the t_ave deviates from 0 for more than one sigma; if only statistical errors are used for the asymmetry, this deviation could be due to errors that are specific for the data set as the error on PbPt, on the dilution factor etc. I would suggest to try to include the errors that are specific for the relevant data set and repeat the evaluation of the t- distribution.
***A***: Yes, we think these deviations are due to the systematic uncertainties. The way we calculate the systematic errors makes it difficult to incorporate the systematic errors into the t-tests. I can use the total sys error determined after the full analysis and redo the t-tests by adding the total sys error to the statistical error in quadrature.


-- another reviewer adds to the point above: otherwise, what are you doing right now of the cases where the t-test fails? Looks like you're claiming the difference is due to the high systematics and still combining those data. I definitely agree with the other reviewer, the systematics should be included when combining the data sets.
***A***: Using the total sys will work for the t-test but it will not work for weighting the data sets while combining them (since the weight from sys error addition will be the same for both sets). Our sys errors have kinematic dependency (a function of q2 and w) but it does not have "data set" dependency.  At this point, we don't have a way to calculate separate sys errors for each separate data set. Only way I can come up for this is adding up the errors from PbPt and dilution factor for each data set in quadrature and add that to the overall statistical error (in quadrature) to use as a weight factor while combining the data sets. This will be quite an effort and may take sometime. (Other sources of the systematic errors are all independent of "data sets", which makes their consideration for this effort unnecessary.)

- section 1.10.2: you refer to the proton note, but could you at least quantify here - or show a plot - the results obtained for the systematics uncertainties?
***A***: email Rob and request plots.


- page 59, last line: "Instead, it was added to the value of the polarization used for the standard analysis, for one data set at a time, keeping others unchanged." In case the error is not statistical, but has a systematic shift (due to procedure, ex. accounting the background) which is the same for all sets, what will be the effect on the overall systematics? The error of 10% on PbPT is so big that all other errors seem to be not as relevant.
***A***: We can't think of any additional source of systematic error on the calculation of PbPt values. Additional 15% systematic error on PbPt would double the overall sys error because of the PbPt. 

- page 60, sentence starting with "For these specific...": after "from the inclusive" you should add the words "and exclusive" or else your sentence makes no sense.
***A***:

- section 1.10.4: how did you pick the modified values for C1 and C2 to use for the error evaluation?
***A***:

- page 61: "For A2(DIS), we used an alternative model with a twist-3 part, g_T^HT added to the standard Wandzura-Wilczek form." Any reference? What is the correction due to the interaction-dependent piece?
***A***:


Physics results

- for the whole "Results" section, we really need to know what plots are meant to be published.
***A***: The draft publication will be attached to the analysis. 

- is the model used to parametrize the structure functions in the analysis the new one resulting from the fit discussed in chapter III?
Reply: Yes.

- in Fig. 18 (and in several other ones in the following), first plot, you show systematic errors in a region where no data are included, I suspect because your data point is out of the y-range of the histogram.
Reply: Yes, I can try to fix it if it is a problem. We will definitely be more careful about these types of issues for the plots we will put into the publication.

- caption of Fig. 21, second sentence: substitute "central" with "green", or else it is wrong. Also, here you say that the biggest error comes from the unknown A2 values, but this error doesn't appear in the set of individual systematics shown in this plots. Or is it included in the "models" error? But then, if this is the case, it doesn't look the major contributor to the uncertainty, as radiative corrections seem slightly bigger, by eye.
Reply: I changed the caption. This caption was transferred from my thesis, when the radiative corrections were smaller and the structure of the figure was different. So, this is an artifact that I should apologize for. I changed the caption to reflect the current situation better in the following way: "{ Virtual photon asymmetry $A_1$ versus $W$ for a few $Q^2$ bins are shown together with systematic errors. The shade at the bottom is the total systematic error. The other systematic errors are offset to the following vertical scales, from top to bottom: pion and pair symmetric contamination (-0.4); dilution factor (-0.6); radiative correction (-0.8); P$_b$P$_t$ (-1.0); models (-1.2); polarized background (-1.4). Uncertainty from the $A_2$ models is a major contributor to the overall systematic error for $A_1$. This systematic error can be reduced once we have measurements on $A_2$.}"

- page 72, Eq.99: What is the point of calculating y using obscure variables like \eta and \epsilon?
Reply: The descriptions here were given for a general reader. The cut was not made by using these variables. Instead, we used p > 0.20 * Ebeam for this cut (for the electron Eout = p for practical purposes).

- page 73, Eq. 100: What justifies the choice of x=0.001 as a lower limit?
Reply: Our model goes to zero at these kinematics.

- caption of Fig. 29: the eg1a data are mentioned, but are not shown in any of the two plots.
Reply: This was a mistake on my part while copying and pasting text from other places. EG1a data is not supposed to be here. I removed it from the caption.

- caption of Fig. 30: the previous analysis shown here is not mentioned anywhere in the text, unless I missed it. Also, here you say that "the two independent analyses complement each other well", while I see that for some Q2 values, such as from 0.35 to ~1 (except one point) the old analysis is consistently below the new one.
***A***: Remove the previous analysis results from here? The previous results refer to the analysis of Vipuli D. using 1.X and 5.X data. Addition of 2.X and 4.X data improved the results in the Q2 region you refereed. Since the models are also improved in this region, it is expected that this analysis results will be an improvement over the previous analysis. 

Modeling the world data

- is the new world data parametrization going to be published together with this analysis result?
Reply: No, that will be a separate publication.

- page 86, related to the previous question on page 61: "The Wandzura-Wilczek relation and the Burkhardt-Cottingham Sum Rule [8] were used to estimate the DIS extrapolation of A2" What are the relative sizes and errors of those corrections based on certain assumptions.
***A***:

- page 89, How does the extracted A_2^n (using the 3He data) compare to known fits in the DIS region? The parametrization by HERMES which was used as DIS prediction was extracted at different kinematic conditions. PDFs seem to have strong Q^2 dependence at small Q^2 and a comparison with other parameterizations will be helpful.
***A***: