Detection anD correction of linguistic errors : results accorDing to linguistic preferences anD uses

This paper examines the ability of 256 university students to detect and correct errors related to different linguistic aspects of Catalan, by means of a test including correct and incorrect sentences. The sentences included aspects of Catalan prescriptive grammar, errors caused by interference from Spanish, and discursive aspects (adequacy, coherence and cohesion). In addition to that, the ability of the students has also been linked to data from their language use and linguistic experience in the following areas: language spoken in the family, language spoken with friends, and formal and informal languages of preference. The results show that students whose preferred language is Spanish or either Spanish or Catalan are less competent than those using Catalan on a regular basis. Taking into account the type of error, students who use preferably Spanish or both languages show significantly worse results in the error detection task (in sentences containing grammar and interference errors) than those who prefer Catalan. In contrast, no significant differences are shown regarding discourse errors when the students’ linguistic use is considered.


intRoDuCtion
The difference between novice and expert writers is a topic that has been studied since the 1970s (Emig, 1977;Flower and Hayes, 1980;Bereiter and Scardamalia, 1987; among others).It is held that expert writers use a series of complex cognitive strategies that, apart from grammar skills, also includes aspects related to Psycholinguistics, Sociology, Pragmatics or Sociolinguistics.
The writing process of expert writers is complex, and it consists of three basic stages: planning, translating (textualization) and reviewing.Each one of these stages involves the knowledge of multiple skills.Text reviewing with regard to both the form (superficial revision) and the content (accurate revision) plays a central role in the writing process (Ferris, 2006;Goldstein, 2006;Sachs and Polio, 2007).In fact, there are many studies that have focused on various aspects linked to text reviewing: from the process of reviewing in general (García and Arias-Gundín, 2004;Pérez, 2001;Conrad and Goldstein, 1999;Chanquoy, 1997;Cameron et al., 1997;Camps, 1992) to specific grammatical features (Largy, Chanquoy and Dédéyan, 2004;Largy, Dédéyan and Hupert, 2004).
But it is obvious that, in order to be expert writers, it is essential for students to be able to review their own texts critically with regard to the content and the text structure as well as the use of language.This means that the reviewing process will involve the knowledge of a series of skills.On the one hand, an accurate reviewing will imply the control of the objectives set, the development of the ideas, and text adequacy, coherence and cohesion, among others.So the correct use of cohesive elements (of text grammar) will play a significant role.On the other hand, one of the main problems with respect to the superficial reviewing has to do with the ability to detect grammatical errors (relative to the prescriptive grammar of the language examined or loan translations borrowed from other languages) and orthographic errors.
From a university perspective, precisely one of the problems observed in Catalan students, and one that causes more concern, is the lack of knowledge of the language rules.In particular, rules related to aspects of Catalan itself and aspects derived from the influence of Spanish, or the weak knowledge of textual elements.Despite all these shortages, Catalan students get results similar to the rest of Spanish students with regard to their linguistic skills, according to the results from the comparative studies on linguistic skills.Those results are obtained when the Catalan in Catalonia is compared to the Spanish in the rest of Spain or when what is exclusively analyzed is the competence in Spanish in both places.Arnau (2004) reviews different studies along these lines (even though he does not specifically measure the detection and correction of errors).
Firstly, the reading comprehension tests from the PISA report (OCDE, 2001) focus on the access and recovery of the information, the integration and interpretation of the information, and the reflection and assessment.The results in Catalonia are always slightly better than the Spanish average, and they are close to the average of the group of countries studied.Those results are maintained in the reports published on the PISA tests in posterior tests (MEC, 2007;MEC, 2010).
Secondly, the reports from the Instituto Nacional de Calidad y Evaluación (INCE, 1999;INCE, 2000) analyze the knowledge of Spanish in Catalonia and in the rest of Spain.They take four dimensions into account: oral and reading comprehension, writing, knowledge of the use of language and Spanish literature.The general results indicate that the knowledge of Spanish in Catalonia is comparable to the average in Spain.Finally, the comparative study on the linguistic skills of students in Catalonia (SEDEC, 1998) compares the results obtained on the knowledge of Catalan and Spanish.It also provides information on the native language of the students (Catalan or Spanish).This study focuses on 14 different dimensions: oral comprehension, morphosyntax, spelling, reading comprehension, writing (adequacy, coherence, cohesion and linguistic correction), oral text (linguistic correction and organization), phonetics and reading (mechanics, intonation and speed).In the assessment of Catalan, the results are always higher for Catalan-speaking students (with statistical significance) or similar to Spanish-speaking students.In the assessment of Spanish, on the other hand, the results are much more homogeneous in both groups.From the 14 dimensions analyzed, there are four with significant differences (in three of which Catalan-speaking students get a higher score and only in one, oral expression, Spanish-speaking students obtain better results).
These results can be explained in the framework of the Interdependence Hypothesis (Cummins, 1979).According to this hypothesis, students are capable of transferring some linguistic knowledge learned in one language to another language.For example, the ability to create coherent, cohesive and adequate texts, or the ability to make metalinguistic judgements, to explain, narrate, etc.
In order to tackle this issue, this article focuses only on the following aspects of reviewing: the ability to successfully detect and correct errors.We present empirical results on this skill from 256 Primary Education undergraduate students (future primary school teachers) in their first year at the University of Girona (during the 2009-2010 academic year).The goal is to analyze this ability and connect it to various data indicating the students' preferences and linguistic uses.
The test provided consisted of one error detection and correction exercise.Following van der Pool's (1996) criteria, who classifies the assessment tools of the reviewing process into online tools (direct observation, interview while the reviewing is being carried out, thinking out loud or specific detection and correction tests) and offline tools (questionnaires, interview after the reviewing task or a posteriori evaluations by the teachers of texts produced by the learners), we can say that our test belongs to the online category, similarly to studies like Cameron et  More specifically, the task designed for this research was based on James (1998) analysis, who divides the learner's linguistic errors into three categories: intralinguistic errors (caused by difficulties of the language itself), interlinguistic errors (due to the influence of other languages) and errors related to discourse properties (including both discourse errors and pragmatic ones).Thus, our test takes these three categories into account and deals with the following types of error: a) errors relative to the prescriptive grammar and orthography specific to Catalan (intralinguistic errors); b) errors related to the linguistic interference of Spanish (interlinguistic errors); and c) errors connected to discourse adequacy, coherence and cohesion (discourse errors).
The results obtained from the participants are connected to their linguistic uses in the following environments: language spoken in the family, language spoken with their friends, preferred formal language and preferred informal language.Hence these data will allow us to know which groups of students have more difficulties detecting and correcting each type of error.

GoALS AnD HYPotHESES
The present study focuses on the empirical analysis of the ability to detect and correct errors related to the prescriptive grammar, to the linguistic interference of Spanish and to discourse elements (of coherence, cohesion and adequacy) and it has the following goals: a.The analysis of how the students' ability to detect and correct errors is connected with their everyday use of Catalan or Spanish.This is done in the following contexts: language normally used in the family; language normally used with friends; preferred language in formal situations; and preferred language in informal situations; b.The finding and analysis of the types of error that offer more difficulties to each group of students.
In b.Students who have Spanish as their preferred language will show a lower ability to detect errors than students who prefer Catalan due to linguistic interference.
c.There will be no statistically significant differences in the effectiveness to detect and correct errors related to Catalan prescriptive grammar, since all the students have had the same formal education in prescriptive grammar during their obligatory education.
d.There will be no statistically significant differences in the detection and correction of errors related to text coherence, cohesion and adequacy, since these are more transferable aspects from one language to another than orthographic or grammatical features.

Participants
The participants in this study were 256 Primary Education first year undergraduate students.The test was answered at the end of the academic year.With regard to their gender, 64% were female and 36% male.The access to university was mainly through a high school diploma and a university entrance examination (70.5%), followed by a vocational qualification (23.4%), examinations for students over 25 years of age (2.8%), and others (from other university degrees or other types of access, 3.3%).The native language of the participants was Catalan (65%), Spanish (25.7%) or both languages indistinctively (9.3%).However, since all participants lived and attended Primary and Secondary schools in Catalonia in their childhood and adolescence, they were supposedly proficient in both Catalan and Spanish.

test design
The test designed consists of 60 sentences.The first 20 sentences are related to the interference of Spanish grammar (interlinguistic errors).The topics selected correspond to those aspects that are more problematic, as identified by Amadeo and Solé (1996) and Mestres et al. (2007), and frequent in the writings of university students, according to the researchers' preliminary observations.Sentences 21-40, on the other hand, deal with grammatical and orthographic issues specific to Catalan (intralinguistic errors).The choice of topics was based on the test of Nivell de suficiència of the Direcció General de Política Lingüística of the Generalitat de Catalunya, which is equivalent to the C1 level of the Common European Framework of Reference for Languages (CEFR).Finally, the last sentences (41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52)(53)(54)(55)(56)(57)(58)(59)(60) have to do with discourse matters (adequacy, coherence and cohesion).
The test started with this statement: "In the following sentences, there can be various problems of language (adequacy, coherence, cohesion, grammar, prescriptive grammar...).Say if the sentence is correct or has problems.If you think that a sentence has problems, write below the version that you consider better or correct."In the appendix we provide the 60 sentences with the particular aspect that is being assessed.The sentences marked with an asterisk are incorrect or have problems.In four questions, more than 50% of the students gave new correct versions of already correct sentences (on average, 64% of the students with Catalan L1 and 61% of the students with Spanish L1).In these cases, we decided to cancel the sentence and not to take them into account | 16 RLyLA Vol.09 (2014), [14][15][16][17][18][19][20][21][22] in the statistical analysis, since the formal context in which the test was handed out could have caused the false impression that some of the sentences contained hidden difficulties.
The reliability of the test has been statistically validated through the split-half method (Spearman-Brown coefficient = 0.791; Guttman Split-Half coefficient = 0.787).
In order to obtain the variables referring to the linguistic preferences, the participants were asked to provide information (through a written form with closed questions) about the language normally used in the family and with friends, and the preferred language in formal and informal situations.Table 1 shows a summary of the participants' characteristics according to their linguistic preferences.As can be observed, there is a clear preference for Catalan in all the environments.The percentage of students who claimed to use other languages (2.5% of the total) was discarded from the statistical analysis since the sample was not representative.Preferred language in informal situations 70.5 16 13.5

Statistical analysis
For the statistical treatment of the data, we used contingency tables (with the specification of the corrected typified residues) and Pearson's chi-square statistic.Once we analyzed whether or not there was a significant relationship between the ability to detect and correct the errors and the factors studied, the next important step was to measure what the direction of that relationship was by comparing the observed cases and the expected cases.The expected cases are the result that would be obtained in the sample analyzed if the two variables (acceptable detection and correction of errors, on the one hand, and language use and preferred language in the various environments studied, on the other) were independent of each other, that is, if there was no relationship between the presence of errors and the factor analyzed.The fact that there is a significant relationship between the two variables causes the observed cases of detection of errors registered not to coincide with the expected ones, precisely because the factor determines that ability.In order to be able to carry out the comparison between cases expected and cases observed, the adjusted residual values (ARV) were utilized.This statistical value follows a normal distribution with zero average and standard deviation 1.Therefore, assuming a confidence interval of 95%, we can say that residual values higher than 1.96 or lower than -1.96 show the situations with the crossing of two variables containing more or fewer cases, respectively, than the ones expected if the variables were independent.

RESuLtS
Firstly, table 2 shows the ARV results, which indicate the ability to detect and correct the errors successfully according to the language used in the various situations examined and the participants' preferred language in the study.The p-value of the chi-square is highly significant in all cases.The results demonstrate that students who claim to use Spanish more than Catalan (or, to a lesser degree, both languages together) have a lower ability to detect and correct errors than students who only use Catalan (since they show significant negative ARV values, while students who claim to use Catalan more often show significant positive values).This aspect needs to be considered since it shows a different profile in those participants who use Spanish on a regular basis.We can see neither a positive transfer in terms of Odlin (1989) nor facilitation in terms of Ellis (1994) and Ellis and Barkhuizen (2005), which would involve an improvement in the learning of the two languages due to their structural similarities, but the performance of Catalan in the students who use Catalan and Spanish or Spanish as their preferred language is less satisfactory than in those students who mainly use Catalan.
| 17 RLyLA Vol.09 (2014), 14-22 Secondly, the analysis of the ability to detect errors according to their typology shows that students who claim to use Spanish more often than Catalan (or the two languages interchangeably) are less able to identify linguistic interference errors and errors specific to Catalan grammar than students who prefer the use of only Catalan.With regard to the detection of errors related to discourse aspects, the groups do not show any differences according to their preferred language.

DiSCuSSion AnD ConCLuSionS
The first goal set in this study was to analyze the ability of different groups of students to detect and correct linguistic errors successfully according to their preferences in the use of Catalan or Spanish and the language normally used in different environments.The observation of the data presented here allows us to state that, in absolute terms, CTR values higher than 1.96 or lower than -1.96 show significant differences.These data agree with the findings provided in the comparative study of the knowledge of Catalan and Spanish sponsored by the Servei d'Ensenyament del Català -SEDEC (1998) (cited in Arnau, 2004), where the knowledge of Catalan and Spanish among 6th-grade schoolchildren in Catalonia is evaluated (this study has the following variables into account: language in the family, socio-professional position of the family and the type of school, i.e., state, private or state-subsidised school).The SEDEC study shows that competence in Catalan (which tests oral comprehension, morphosyntax, orthography, phonetics, reading mechanics, intonation and reading rate) is significantly better in Catalan-speaking students than in Spanish speakers in 9 out of 14 tests.In addition to this, it also clearly demonstrates that the pupils' family language does not cause any setback in the knowledge of Spanish.We can see this in the fact that there are no significant differences in most of the tests, and that only in one of the 14 tests Spanish-speaking students obtain better results than Catalan-speaking students (against three in which Catalanspeaking students get a better score).Hence, the results obtained in our study are in the same line as the results shown by the SEDEC study.Therefore, hypothesis a) is confirmed.
| 18 RLyLA Vol.09 (2014), [14][15][16][17][18][19][20][21][22] As far as the second goal is concerned (i.e., the analysis of which typology of errors shows more difficulties for each group of students according to their linguistic uses), it can be observed that students that prefer Spanish get worse results with regard to the detection of errors caused by interference processes (in a high degree) and errors related to Catalan prescriptive grammar (to a lesser degree, although it is also statistically significant).These data allow us to validate hypothesis b), but not hypothesis c), since the worst results are maintained in the group preferring Spanish in the various situations examined.
Related to the latter hypothesis, the results obtained are not those expected as the aspects on normative grammar that are assessed are specific to the written language and to a formal register (therefore, quite far from colloquial oral uses).Thus, the use of Catalan in the family should not be an advantage.On the other hand, no significant differences are seen on the detection of problems relating to the discourse, so the results validate hypothesis d).
It is interesting to emphasize that the results show a lack of ability to detect orthographic and grammatical errors, interference errors and discourse errors in all groups, although mainly in the group where Spanish is highly preferred.This inability prevents the students from getting the skills in writing that are characteristic of expert writers, among other things because they neglect one of the three fundamental stages in the writing process, i.e., correction.Students who prefer Spanish get results that are overall worse than students who prefer Catalan, so they will need strategies that allow them to improve their communicative competence.
Finally, it has to be stressed that, according to the SEDEC (1998) study, Catalan students (both Catalan speakers and Spanish speakers) obtain general results on linguistic competence in Spanish similar to those observed in Spanish monolingual autonomous communities, but, in the former case, students also gain linguistic competence in Catalan (clearly higher in Catalan speakers).The SEDEC data also indicate that in practically all the tests, both in Catalan and in Spanish, Catalan speakers get better results than Spanish speakers.
The analysis of the students' linguistic skills is a topic that needs to be investigated in more detail.This study contributes in this direction by showing that the ability to detect errors in the Catalan language is higher in Catalan speaking students even when all the students received the same education.Now a matter for further research would be to carry out the test the other way around so we could observe the ability of Catalan speakers and Spanish speakers to detect errors in Spanish.
If the results followed the same direction as our study has pointed out, it would seem that, despite having a good bilingual or multilingual education, the proficiency of the preferred language in the family and with friends would be higher than in other languages.In this case, it would be necessary to find strategies to correct these inequalities in the knowledge of one language so the speakers who do not have it as their preferred one could overcome this disadvantage (which, as we have seen, it is maintained at university) without affecting the knowledge of other languages.This would help us to get an overall view of the situation, and it would also mean a step closer towards understanding how the different skills are transferred from one language to another and how one language interferes with another.
order to formulate the hypotheses, we have considered the results from different evaluation studies (such as SEDEC, 1998; INCE, 1999; INCE, 2000; OCDE, 2001; MEC, 2007; MEC, 2010).So the hypotheses that are assessed here are the following: a. Overall, students who prefer and use Catalan more often will obtain better results in detecting and correcting errors (presented in the Catalan language).

Table 1 .
Percentage of participants according to their preferences and linguistic uses.

Table 2 .
Ability to detect and correct the error successfully according to the preferred language and use.Adjusted residual values and Pearson's chi-square.Asterisks show significant ARV values.

Table 3 .
Ability to detect and correct errors successfully according to the type of error and the preferred language of use in several situations.CTR and indication of the degree of significance of p.