Can You Identify These Celebrities? A Network Analysis on Differences between Word and Face Recognition

Round 1

Reviewer 1 Report

Introduction
The authors must carefully check English grammar.

Keywords: face recognition; word recognition; masked priming; network analysis; celebrities
The authors must carefully to think and to add a few keywords corresponding to the leading Journal's aims and scopes.
The scope must meet the Journal's scopes.

Authors should provide references in square brackets. A style like the references provided by the authors in this article is inappropriate for the Journal of Mathematics. Use the notation [1,2] instead of (1,2).

72-73 lines. From this, Sunday et al. (2019) concluded that geographical places of origin and differences between men and women interact to predict face recognition ability.
The authors use not proper style to refer. Besides, in the list of references, there are two papers by Sunday et al. (2019).

125 line: software (Forster and Forster, 2003) was used.
The authors use not proper style to refer.

The authors should not add references in the last part of the paper (Conclusions). Unfortunately, the authors in this part added 14 new references. These references must be in other parts of the article or deleted from the paper (In such a case, there will be fewer references).

95-97 lines: A total of two experiments, with two different samples and subsamples, were carried out. In the first one, 20 Spanish University students (10 men and 10 women), with no history or evidence of neurological or psychiatric disease, volunteered to participate.
98-101 lines: Secondly, a sample of 60 participants volunteered to take part in Experiment II: a total of 20 Spanish University students (4 men and 16 women), 20 Brazilian University students (6 men and 14 women), and 20 North American University students (2 men and 18 women) volunteered to participate.

Table 2. Mean and SD (standard deviation), and accuracy rate for Experiment II, across conditions.
Please, explain why such accurate results (accuracy 0.01 per cent). According to reviewers opinion accuracy is not more than 1 per cent, or less, because sample size, as well as, the population size of precedence are too small for such accuracy.

The authors should understand that not all people in the world speak and write beautiful Spanish. Therefore, if possible, the bibliography should be written in English in this article, as the MDPI prints this Journal in English.

Author Response

We would like to thank the reviewer for his/her invaluable help and comments.  We believe the paper has been significantly improved. Please, find the new version attached and changes highlighted in red ink. We tried to follow all his/her commentaries, described as follows:

Response to reviewer 1.

Point 1. Introduction
The authors must carefully check English grammar.

An English native speaker has reviewer the manuscript. We thank her, and we have added her name in acknowledgements.

Point 2. Keywords: face recognition; word recognition; masked priming; network analysis; celebrities
The authors must carefully to think and to add a few keywords corresponding to the leading Journal's aims and scopes.
The scope must meet the Journal's scopes.

Some keywords were excluded and some other added in other to better contextualize the current work in the journal's scope. These are described as follows: structure of networks; language networks; network analysis; face recognition; word recognition; masked priming.

Point 3. Authors should provide references in square brackets. A style like the references provided by the authors in this article is inappropriate for the Journal of Mathematics. Use the notation [1,2] instead of (1,2).

We are really sorry for this mistake, as we misunderstood the citation style method in zotero. We hope it is adequate now.

Point 4. 72-73 lines. From this, Sunday et al. (2019) concluded that geographical places of origin and differences between men and women interact to predict face recognition ability.
The authors use not proper style to refer. Besides, in the list of references, there are two papers by Sunday et al. (2019).

We are really sorry for this mistake. We have changed it, in accordance with the reviewer suggestion. 

Point 5. 125 line: software (Forster and Forster, 2003) was used.
The authors use not proper style to refer.

Once again, we are really sorry for this mistake. We have changed it, in accordance with the reviewer suggestion. 

Point 6. The authors should not add references in the last part of the paper (Conclusions). Unfortunately, the authors in this part added 14 new references. These references must be in other parts of the article or deleted from the paper (In such a case, there will be fewer references).

We have created two different sections for discussion and conclusions. This was in accordance with previous literature, e.g.:

Druică, Elena, et al. "Exploring the Link between Academic Dishonesty and Economic Delinquency: A Partial Least Squares Path Modeling Approach." Mathematics 7.12 (2019): 1241.

In this way, any reference was taken out from the conclusion section, as indicated by reviewer 1.

Point 7. 95-97 lines: A total of two experiments, with two different samples and subsamples, were carried out. In the first one, 20 Spanish University students (10 men and 10 women), with no history or evidence of neurological or psychiatric disease, volunteered to participate.

This paragraph was reformulated, in order to make it more straightforward.

 
Point 8. 98-101 lines: Secondly, a sample of 60 participants volunteered to take part in Experiment II: a total of 20 Spanish University students (4 men and 16 women), 20 Brazilian University students (6 men and 14 women), and 20 North American University students (2 men and 18 women) volunteered to participate.

This paragraph was simplified.

Point 9. Table 2. Mean and SD (standard deviation), and accuracy rate for Experiment II, across conditions.
Please, explain why such accurate results (accuracy 0.01 per cent). According to reviewers opinion accuracy is not more than 1 per cent, or less, because sample size, as well as, the population size of precedence are too small for such accuracy.

We have homogenized table 1 and 2, avoiding very low values, that might not make sense, as pointed out. Even if smaller samples are used in Psychology, and we tried to ensure quality of data by an experimental design based on mixed measures that included Repeated measures design. This reduces the variance of estimates, allowing statistical inference to be made with fewer subjects. Moreover, under this design, all participants receive every condition. Nevertheless, we agree the sample size is small. For this reason, two strategies were followed: i) we have simplified the previous network analysis, in order to gain statistical power (for this reason, the second network was smaller than the first one in the previous paper), ii) a bootstrapping technique was employed.

As indicated in the paper, the second strategy seems quite appropriate to the key concerns, by resample an original sample. Therefore, a bootstrapping technique was employed, by setting it of a size N to 1000. We believe this is an added value fir the manuscript.

Point 10. The authors should understand that not all people in the world speak and write beautiful Spanish. Therefore, if possible, the bibliography should be written in English in this article, as the MDPI prints this Journal in English

We have checked the references section. In the new version, there is only one reference which is offered in both Spanish and English. The rest of references are written in English.

We tried our best to follow all the raised point, which we find very interesting for the improvement of the current manuscript. Just in case, if we miss something, we would like to apologize with reviewer 1 in advance, as English is no our mother tongue.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is very interesting and well written (although some moderate English language revisions are required). One of my concerns is that the authors applied ANOVA in the first experiment with only 20 participants.

Given the small sample size, a non - parametric method would have been recommended. The results presented on Lines 177 - 178 may not be reliable - so I suggest a double check of this result using the Kruskal Wallis test. There is agreement among statisticians that ANOVA is robust to many violations of its assumptions, but sample size is a sensible source of serious flaws. 

The second concern is that the samples involved in the second experiment are highly unbalanced in terms of gender. The authors mention (lines 182 - 183) that gender is not a control variable in this second study. This decision is made based on the result obtained in the second experiment and stated on lines 175 - 176: "No statistical differences were found across participants sex or stimuli gender.". I am not sure that this sentence is phrased correctly, but I understand the idea.

However, if the results of the first experiments are not reliable due to the small sample size, and given the unbalanced structure of the samples involved in the second experiment, I am wondering whether the conclusions of the second study are reliable. 

Another problem arising from the size of the samples is that any ANOVA test is reliable if the variance of the data across groups is the same. Let aside that the authors did not check for this condition, I am wondering how the condition can be checked in each case and how many participants would be considered in each sample.

Lines 95 - 101: the authors present the samples, while lines 174 - 175 explain that the results are reported based on gender. What I don't understand is related to Table 1: the first column refers to "Condition", and according to the reporting there are 4 conditions. How many participants (out of the 20 in total) were assigned to each condition? The same applies to Table 2 and the corresponding first column: how many participants were in each condition, and how were they assigned to each group?

According to Table 3, results have been reported within the country - condition contingency: once again, I am concerned with the small number of participants in each group, so I recommend non - parametric methods to conduct the analysis.

I am not familiar with network regression, so I cannot give any feedback in this respect. 

Beside the use of non - parametric techniques to double check the results, I recommend a word of caution in respect to the existing conclusions. Given the samples size in the first experiment and the unbalanced gender distribution in the second experiment, they could be too far reaching. Maybe the authors could include a paragraph where the main limits of the research should be clearly stated. 

Overall, the paper is promising and, if revised, I believe that it will deserve publication. 

 

Author Response

Dear Reviewer 2,

We tried to follow all your suggestions. Thanks to that, we believe the paper has been significantly improved. Please, find the new version attached as a new main document with changes highlighted in red ink. Thank you again for your help and inestimable time.

 

RESPONSE TO REVIEWER 2

Point 1. The paper is very interesting and well written (although some moderate English language revisions are required).

Thank you, these words are very constructive for us. An English native speaker has reviewed the manuscript. We thank her in the Acknowledgement section.

Point 2. One of my concerns is that the authors applied ANOVA in the first experiment with only 20 participants. Given the small sample size, a non - parametric method would have been recommended. The results presented on Lines 177 - 178 may not be reliable - so I suggest a double check of this result using the Kruskal Wallis test. There is agreement among statisticians that ANOVA is robust to many violations of its assumptions, but sample size is a sensible source of serious flaws.

We would like to thank the reviewer 2 for these commentaries. We have followed the suggestion, which we find has added value to the current manuscript.

Point 3. The second concern is that the samples involved in the second experiment are highly unbalanced in terms of gender. The authors mention (lines 182 - 183) that gender is not a control variable in this second study. This decision is made based on the result obtained in the second experiment and stated on lines 175 - 176: "No statistical differences were found across participants sex or stimuli gender.". I am not sure that this sentence is phrased correctly, but I understand the idea. However, if the results of the first experiments are not reliable due to the small sample size, and given the unbalanced structure of the samples involved in the second experiment, I am wondering whether the conclusions of the second study are reliable.

We would like to specially thank the reviewer again, as this interpretation was addressed in a biased manner in the prior manuscript. We tried to be more cautious in our affirmation regarding sex effects. Moreover, sample size has been highlighted as unreliable for differences across mean and women.

Point 4. Another problem arising from the size of the samples is that any ANOVA test is reliable if the variance of the data across groups is the same. Let aside that the authors did not check for this condition, I am wondering how the condition can be checked in each case and how many participants would be considered in each sample.

We have reported Levene test, as well as normality tests. We have employed a similar sample size from previous literature on Priming effects:

Gil-López, C., Perea, M., Moret-Tatay, C., & Carreiras, M. (2011). Can masked priming effects be obtained with words?. Attention, Perception, & Psychophysics, 73(6), 1643.

Perea, M., Comesaña, M., Soares, A. P., & Moret-Tatay, C. (2012). On the role of the upper part of words in lexical access: evidence with masked priming. Quarterly journal of experimental psychology (2006), 65(5), 911-925.

 

GPower was also employed to explore this issue, which was pointed out by all reviewers. Regarding Experiment I, there is a 92% chance of correctly rejecting the null hypothesis of no significant effect of the interaction with 10 males and 10 females for a total of 20 participants. With regards to experiment II, Regarding Experiment I, there is a 88% chance of correctly rejecting the null hypothesis of no significant effect of the interaction for a number of 28 measures across 3 number of groups (Please, see attached document).

We hope this could justify our parametric approach. Nevertheless, we have addressed in a non-parametric way as well. We believe this is an added value for the paper.

Point 5. Lines 95 - 101: the authors present the samples, while lines 174 - 175 explain that the results are reported based on gender. What I don't understand is related to Table 1: the first column refers to "Condition", and according to the reporting there are 4 conditions. How many participants (out of the 20 in total) were assigned to each condition? The same applies to Table 2 and the corresponding first column: how many participants were in each condition, and how were they assigned to each group?

We employed a Repeated measures design (within between sex interaction for Experiment I and Country interaction for Experiment II). This reduces the variance of estimates, allowing statistical inference to be made with fewer subjects. Moreover, under this design, all participants receive every condition. As stimuli are divided in 50% of female Celebrities, and 50% of male celebrities, differences across sex participants were described in table 1. We have reformulated the labels in table 1, as these were not clear.

As highlighted by the reviewer, there are 4 conditions (Identity Priming, related Priming, Unrelated priming in same nature of stimulus, Unrelated priming in different nature of stimulus,) x 2 nature of stimulus (faces/names) x Block (Target/Distracting). We have also reformulated the Method section to clarify this issue, which was kind of blurring in the previous version.

Point 6. According to Table 3, results have been reported within the country - condition contingency: once again, I am concerned with the small number of participants in each group, so I recommend non - parametric methods to conduct the analysis.

We have implemented previous analysis with non-parametric methods. These are of interest, as suggested by reviewer 2, as these provide similar results from the parametric approach.

Point 7. I am not familiar with network regression, so I cannot give any feedback in this respect.

In order to reach all audiences in the readership of this journal, we have added a friendly tutorial paper on network analysis for Psychology in the references. We hope this could be of interest. We have simplified the networks in terms of nodes to gain more statistical power, and we have added a bootstrapping technique to resample an original sample, as indicated by reviewer 1 and 3.

Point 8. Beside the use of non - parametric techniques to double check the results, I recommend a word of caution in respect to the existing conclusions. Given the samples size in the first experiment and the unbalanced gender distribution in the second experiment, they could be too far reaching. Maybe the authors could include a paragraph where the main limits of the research should be clearly stated.

We have included this in the discussion section. Thank you for this crucial remark.

Point 9. Overall, the paper is promising and, if revised, I believe that it will deserve publication.

We really thank the reviewer 2 for such a motivating commentary. We tried our best to follow all the raised point, which we find very interesting for the improvement of the current manuscript. Just in case, if we miss something, we would like to apologize with reviewer 2 in advance, as English is no our mother tongue.

Author Response File: Author Response.pdf

Reviewer 3 Report

Considering my area of expertise, I could check the analysis section. It sounds fine and original.

Moreover, as a researcher I doubt the number of samples (from 20 and 60) participants is sufficient.

Author Response

Dear Reviewer 3,

We tried to follow  your suggestions, that make us believe the paper has been significantly improved. Please, find tthe new document with changes highlighted in red ink. Thank you again for your inestimable time and time to help us to improve the manuscript.

 

RESPONSE TO REVIEWER 3

Considering my area of expertise, I could check the analysis section. It sounds fine and original. Moreover, as a researcher I doubt the number of samples (from 20 and 60) participants is sufficient.

Thank you for pointing this out. We have indicated previous literature where smaller samples are used in Psychology. This is not a justification,for the current study but a common practice in the field. In this way, we tried to ensure quality of data by an experimental design based on mixed measures that included Repeated measures design. This reduces the variance of estimates, allowing statistical inference to be made with fewer subjects. Moreover, under this design, all participants receive every condition.

Nevertheless, we agree the sample size is small. For this reason, several strategies were followed: i) we have employed both parametric and non-parametric approaches that showed similar results, ii) we have simplified the previous network analysis, in order to gain statistical power (for this reason, the second network was smaller than the first one in the previous paper), iii) a bootstrapping technique was employed.

As indicated in the paper, the second strategy seems quite appropriate to the key concerns, by resample an original sample. Therefore, a bootstrapping technique was employed, by setting it of a size N to 1000. We believe this is an added value for the manuscript.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

No more comments or concerns.

Reviewer 3 Report

My concerns are addressed in the revised verson.