English & Armenian Dictionary 2.7 serial key or number

• Magdalena Łuniewska, 
• Zofia Wodniecka, 
• Carol A. Miller, 
• Filip Smolík, 
• Morna Butcher, 
• Vasiliki Chondrogianni, 
• Edith Kouba Hreich, 
• Camille Messarra, 
• Rogayah A. Razak, 
• Jeanine Treffers-Daller

Age of acquisition effect

Age of acquisition (AoA), i.e. an estimation of the age at which a word is acquired, is an important factor in research on word processing. A growing number of studies have shown AoA to affect performance in various psycholinguistic tasks, such as picture naming, word naming or lexical decision [1–3]. In particular, words acquired earlier in life are processed faster and more accurately than words learned later. This phenomenon is called “Age of acquisition effect”. Recently, AoA effect has been replicated in lexical decision tasks in English [4], Dutch [5]and Persian [6], in a semantic decision task in German [7], in free word recall tasks in Russian [8], in recognition memory tasks in English [9,10], as well as in picture naming and word recognition in both bilingual [11,12] and monolingual children [13].

The AoA effect has recently been also found in new tasks such as intentional and incidental forgetting [14], elicitation of tip-of-the-tongue experiences [15], as well as eye movement in reading [16]. In particular, early acquired words are more resistant to being forgotten, even if participants are instructed to forget the list of words [14]. Consequently, early words appear less often in tip-of-the-tongue phenomena, when a participant knows a word but cannot produce it [15]. Early words are also read faster, and lead to shorter fixation durations during novel reading [16]. Similarly to single words, an AoA effect has recently been shown in processing of multiword phrases: early acquired multiword phrases are processed faster [17].

AoA ratings are negatively correlated with word frequency, i.e. the higher the word’s frequency, the earlier it is acquired [18–27]. However, previous research has shown that the effects of AoA and word frequency are not identical and AoA effects may also be registered when word frequency is controlled in such tasks as picture naming or lexical decision [28–36]. On the other hand, a words’ AoA may depend on its frequency trajectory, i.e. the distribution of experiences of the word in time [23,37–39]. For that reason, authors suggested the use of frequency trajectory instead of AoA estimations to examine learning effects in lexical processing [38]. Nevertheless, application of either word frequency or frequency trajectory in cross-linguistic studies may be impossible, as finding comparable sources of word frequency in more than two languages may be challenging, and for many languages we still lack frequency data. In general, frequency data may be obtained in a language for each single word, when big enough corpus is available. This is not the case of subjective AoA ratings, which may be collected only for a limited set of words [40], even if the set of words is relatively big up to several thousand words [22,40–49]. However, in contrast to frequency data, AoA ratings may be collected in the same manner across a wide range of languages and cultures [1]. The difficulty with comparability of available sources of word frequency across languages and the independence of AoA and frequency effects, bring out the need for a cross-linguistic parallel database of AoA ratings.

Order of word learning across languages

Although it is typically assumed that pace of vocabulary learning is similar across languages (despite acknowledging significant individual differences among children [50]), this assumption has not been studied systematically across a wide range of various languages [13].The pioneering research on modelling the order and pace of early word acquisition across seven languages is currently conducted with the use of the MacArthur-Bates Child Development Inventories (CDI) database [51]. First reports from this study suggested that the order of early acquisition of vocabulary may be predicted by the use of similar variables across languages, such as word frequency, concreteness and ‘babiness’ [51].

There are two basic methods of estimating at what age words are learned: an objective method based on wide studies of children acquiring the language [18,26,27,40,52–56], and a subjective one based on adults’ estimations of when they learned the words [1,19–21,41,44–46,49,53,57–96]. Both subjective and objective estimations of the age of word acquisition has been studied across several languages. The list of languages studied has been reviewed in detail [1] and includes such languages as Chinese, Dutch, English, French, German, Greek, Icelandic, Italian, Japanese, Norwegian, Persian, Portuguese, Russian, Spanish and Turkish. Recently new subjective estimations of AoA have been published for Arabic [97], German [47], Italian [98], Polish [48], Spanish [99,] and Turkish [] words. Though objective AoA ratings are, by definition, more valid than subjective estimations, they are much more difficult to obtain, as collection of objective AoA data involves studies of huge samples of children [1,18,26,27,40,55,56]. On the other hand, subjective AoA estimations are highly correlated to the objective measures [1,40,53,,].

Although either subjective or objective AoA data have been published for a wide list of languages, a direct comparison of the AoA estimations in two or more languages may be challenging, as usually the studies differ in terms of the list of items included in the study and in terms of the measurement scale used [1]. So far, the only study which employed the same method of data collection across 25 various languages from five language families was [1] and the current study broadens it with seven new languages. The previous study found moderate to high correlations of order of words between all language pairs under scrutiny. Existence of such correlations was earlier postulated [18,], though the analysis was based on data collected in a various ways for the languages.

The current study

Here we provide new AoA ratings for seven languages: American English (a Germanic language spoken in the US by million people []), Czech (a West Slavic language spoken mostly in the Czech Republic with millions of native speakers []), Gaelic (a Celtic language spoken as a minority language in Scotland by 57, people []), Lebanese Arabic (a dialect of Arabic spoken primarily in Lebanon by 94% of the Lebanese population though there are no precise estimations of its number []), Malay (a major language of the Austronesian family spoken by 19 million speakers in the world with million speakers in Malaysia []), Persian (standard Iranian Persian spoken in Iran as the official language by 50 million people [] in contrast to other variants of Persian spoken by milion people worldwide []) and Western Armenian (one of the two standard forms of the Armenian language, spoken worldwide in the Armenian diaspora by million people [], outside the Republic of Armenia; in the current study all participating speakers of Western Armenian lived in Lebanon). For five languages included in the current paper (except for American English and Persian) no previous AoA data were available so far, which enhances the novelty of the current study.

We used a homogenous method of data collection in order to obtain ratings comparable across the seven languages studied in the current paper. Because we used the same method as in the previous study in 25 languages [1], the current paper results in a comparable database of AoA ratings in 32 languages (see S1 Table). In addition to providing the new database of AoA ratings, we asked the question of whether the previously found pattern of cross-linguistic similarities in the order of word acquisition occurs also among the languages not included in the previous research. By applying hierarchical clustering, we explored the patterns of similarities of AoA ratings across languages. Across the seven languages, we searched for predictors of AoA estimations by comparing the words’ AoA ratings with words’ length and frequency. Finally, in order to characterize the early vocabulary across languages, in each language we examined whether the first and the latest words differ in terms of length in syllables and phonemes, as well as in the number of compound words.

The database of AoA ratings across 32 languages which came from the original set of 25 languages [1] combined with the current study may be of particular use in cross-linguistic studies, where comparable data for several languages are needed. For instance, AoA collected in the same way across languages may be applied as predictor in studies of vocabulary acquisition [11–13] or multilingual studies of lexical decision [6].

Participants

Research was approved by Ethics Committe of Faculty of Psychology, University of Warsaw. Participants were adults, with a minimum of 21 per language (total range: 21 to 37, M = , SD = ; see Table 1). Participants were females (69%) and 54 males (31%), aged 18 to 68 (M = , SD = ). The group was unbalanced in terms of gender, however the previous study on AoA across 25 languages reported no significant difference in the mean ratings between the women and men [1]. Participants were recruited via academic communication (university subject pool or lecturers informing students about the study), social media and through neighbourhood networks as well as chain-referral sampling. Data were collected in the US (Pennsylvania, for American English); Czech Republic (for Czech); UK (Scotland for Gaelic), Lebanon (for Lebanese and Western Armenian); Malaysia (for Malay) and Iran (for Persian). Participants took part in the study voluntarily, and in the case of American English and Czech received course credits. In the case of Gaelic, participants obtained vouchers for online shopping. All participants reported their gender, age, education level, occupation, country of residence, native language, number of spoken and used languages, and number of they have. In all languages participants additionally reported the age of their children, and in Gaelic some additional questions regarding the current and previous use of languages were asked. All participants provided AoA ratings for all words included in the survey.

All participant demographic data as well as their responses are available from: manicapital.com (DOI: DOI /manicapital.com).

Stimuli

The same sets of nouns and verbs (a total of words) were used in each language. In each language, the lists of exact word forms were obtained in a naming study. However, the procedure of selection of exact word forms differed between the languages.

In all languages except Lebanese Arabic, i.e. American English, Czech, Gaelic, Malay, Persian and Western Armenian, the lists of items were obtained in separate naming studies which used the same set of pictures drawn for Cross-Linguistic Lexical Tasks (LITMUS-CLTs, []). In the naming studies, a minimum of 20 native speakers of the target languages provided labels for the pictures. The most frequently given labels for each of pictures were then used as target items in the current AoA study. In several cases, the naming procedure brought no evident dominant name for a picture but several different names were provided with similar frequency. For these cases the additional words (mostly synonyms of the target words) were added to the list applied in the AoA procedure for a given language. Thus the exact number of items in each language was: nouns + verbs for American English, nouns + verbs for Czech, nouns + verbs for Gaelic, nouns + verbs for Malay, nouns + verbs for Persian, and nouns + verbs for Western Armenian. In order to enhance the comparability of the ratings across the languages, only the ratings for the dominant names (the one most frequent label for a picture in a given language) are reported in the current paper and used in the statistical analyses (a total of words, the same as used in the previous study).

In the case of Lebanese Arabic, the items were selected in the same manner as in the 25 languages reported in the study by Łuniewska and colleagues [1], following a previous online picture naming study conducted in 34 languages, aimed at selecting target words for Cross-linguistic Lexical Tasks, LITMUS-CLTs [].

Although the exact lists of word forms resulted from the picture-naming procedures, in the current study only words were used as stimuli, without any pictures. In each of the languages both the picture naming study (going beyond the purposes of the present paper) and the AoA study (described here) were preparatory studies for an ultimate goal of designing Cross-linguistic Lexical Tasks for these languages.

To access the relationship between AoA and word characteristics for each language, we used three measures: (1) length in phonemes, (2) length in letters (Table 2), and (3) word frequency. In case of word frequency, data were available for four of the languages only: American English [], Czech [], Malay [] and Persian [96] (only noun frequency).

Procedure

The same instructions as used in [1] were applied in all languages but Gaelic. All materials (the website, instruction, examples etc.) were translated from English into each of the languages involved by native speakers who were also researchers (linguists or psycholinguists).

The procedure was available online via a website designed exclusively for the purposes of the study (manicapital.com). The website was made available in all 32 languages (new languages and languages studied previously by [1] therefore participants could use their native language exclusively while using the website. After entering the website, participants were instructed to download a file and open it in Microsoft Excel (or Open Office). The file contained four sheets. The first sheet presented basic information about the study and the instructions, and the second sheet contained questions on the demographics of the participants. The lists of nouns and verbs were presented on the third and fourth sheets, respectively. All the instructions, questions and words were presented in the mother-tongue of the participants, i.e. the language of assessment. The original spreadsheets used in the study are available from: manicapital.com (DOI: DOI /manicapital.com).

Participants were asked to estimate at what age they had learned the words presented in the two sheets. The exact form of the instruction was: “For each word please estimate the age (in years) at which you think you learned this word; that is, the age at which you would have understood that word if somebody had used it in front of you, even if you did not use, read or write it at the time”. The exact form of the question was: “When did you learn the word?”. Participants were asked to type a number from 1 (if they thought they had learned the word when they were one year old) to 18 (if they thought they had learned the word when they were 18 or older). They were encouraged to guess the age if they were not sure and not to spend too much time on any single word. If they did not know the word, they were asked to enter “X” in the box.

To ensure that the participants understood the instructions, we provided four examples of both nouns and verbs acquired early and later in life. The examples were presented in a table that looked similar to the one filled out by the participants. Explanatory comments were added to the table (e.g. “Someone estimates that s/he learned the word ‘to ask’ at the age of 3 years.”).

The words on both the noun and the verb list were presented in random order, generated individually for each participant during the file downloading. In the Nouns and Verbs sheets, below the list of words, a short thank-you note was presented together with a reminder of the other sheet (“Thank you for filling in the table for nouns. Have you filled the table for verbs as well?”). Each participant was given the full list of words. Task duration was about half an hour. After filling in the file, participants were asked to upload it via the website or send it as email attachment to the address reserved for the purposes of the study.

In Gaelic, very similar procedure was used. The only difference was that instead of a downloadable Excel file, an online Qualtrics survey was used. Gaelic participants answered the same question (translated to Gaelic) as speakers of other languages, and used the same scale to give their responses. The noun and verb parts were separated and Gaelic participants answered the whole set of questions. A Gaelic version of the spreadsheet is also available from : manicapital.com (DOI: DOI /manicapital.com).

Statistical analyses

We first calculated average age of acquisition of each word in each language. Then for each participant we calculated Pearson’s correlation coefficients between the average AoA in his/her language and his/her estimations. Next, for each language we calculated the average value and standard deviation of the correlation coefficients. We aimed at excluding data from participants whose correlation coefficients would be lower than language average minus 2SD, though there were no such participants and thus data from all participants were included in further analyses.

For each word in each language we calculated the average age of acquisition. In order to check the reliability of the data, we computed split-half reliability coefficients in each language. To compute the coefficients, we randomly divided participants of each language into two subgroups and calculated the Pearson’s correlation coefficients for the average AoA ratings provided by the two groups.

To verify cross-linguistic patterns of word order, we computed Spearman’s rho correlations between average AoA in each of the seven languages reported in the current study and 25 languages reported in [1]. Then we ran a hierarchical clustering of variables, i.e. languages, with the use of R package ClustOfVar [] in order to explore the patterns of similarities of AoA ratings across languages. In this package, distances between variables and clusters are based on correlation coefficients.

To examine the relationships between words’ AoA and their other characteristics in each language we calculated Spearman’s rho correlations between average AoA and word length in phonemes and letters, as well as frequency, if there were applicable data. Then we selected the earliest and the latest 30 words (up to 37 if there were more words of the same AoA) in each language and compared the length in phonemes and letters of these words with the use of a series of t-tests. Additionally, we compared the number of compound words between the earliest and the latest words with the use of a series of chi-square tests. The exact values of stimuli characteristics (word length in phonemes and letters, information about compound words and word frequency) are available as S2 Table.

Reliability of the data

The split-half reliability coefficients were very high and ranged between and (S3 Table).

Descriptive results

The average AoA ratings obtained for each of the seven languages are presented in the supplemental material. The vast majority of the words, namely 94% of the words, were assessed as known before the age of 7 years (Fig 1) and the remaining 6% of the words were estimated to be acquired between the age of 8 and 13 years.

Fig 1. Means for AoA ratings across seven languages.

The dots represent the words which are outliers. The solid line represents the mean for the seven languages (M = ), the dotted line represents the mean for 25 languages (M = ) studied in [1].

manicapital.com

Cross-linguistic comparison

The ratings in the seven languages were moderately to strongly correlated with each other. Among the seven languages Spearman’s rho adjusted for split-half reliability varied from for Gaelic-Malay pair to for Western Armenian-Lebanese Arabic pair (see S3 Table and Fig 2). The ratings in the seven languages were also moderately to strongly correlated to previously collected ratings in [1]. For the seven languages contrasted to the 25 languages included in the previous study, Spearman’s rho adjusted for split-half reliability varied from for Malay-Hungarian pair to for Czech-Slovak pair, with mean and median for all languages equalled and (S3 Table and Fig 2).

Interestingly, while analysing the entire dataset (including the previously examined 25 languages and the added seven languages), we found a consistent pattern of correlations: (1) the languages close in terms of family had the highest correlation coefficients (American English vs. British English: , vs. South African English: ; Czech vs. Slovak: , vs. Polish: , all being West Slavic languages; Lebanese vs. Maltese: , both being Semitic Languages; Gaelic vs. Irish: , both being Celtic languages); (2) Malay (a language from an Austronesian family not closely related to any other languages in the sample) had relatively low correlation coefficients (mean: ), comparable with the mean correlation for Hungarian (mean: )–the language least similar to others included in the previous study [1], in terms of both origin and correlation coefficients obtained here; (3) Western Armenian (assessed in the Armenian minority in Lebanon) was most strongly related to Greek (a language said to be historically close to Armenian []): and Lebanese (in the case of this study close not by language family but by territory): ; (4) Persian had the strongest correlations with Turkish: , Maltese: , and Greek:

These patterns were mostly confirmed with the hierarchical clustering (Fig 3). In particular, in the hierarchical clustering we found that: (1) the three closest languages, forming the first cluster, were South African English, American English and British English, (2) most Germanic languages formed a joined cluster with Finnish, spoken in similar geopolitical settings, (3) Czech, Slovak and Serbian, all being Slavic languages, formed a joined cluster, (4) IsiXhosa and Malay, the single representatives of the Bantu and Austronesian languages, joined the clustering at the highest distance, i.e. were the least similar languages to all others. There are however some exceptions to this general pattern which should be noted: e.g. Polish is not clustered with other Slavic languages but with Afrikaans, a Germanic language, which in turn is not clustered with other Germanic languages.

AoA and word characteristics

In all languages except Lebanese Arabic, both word length in phonemes and word length in letters were positively correlated to average AoA (Table 3). The significant correlation coefficients varied between low ( – in Western Armenian) to moderately high ( – in American English and – in Persian). Additionally, in all languages with existing frequency data we obtained moderately high negative correlations of AoA and word frequency (Table 3).

In all languages except Lebanese Arabic we found significant differences between the length of the earliest and the latest words in terms of both phonemes and letters (Table 4). The average difference between the earliest and the latest words’ length was phonemes and letters.

In two languages, Gaelic and Persian, we found a difference in compound word distribution between the earliest and the latest words (Table 4). In these two languages, there was significantly more compound words among the latest words than among the earliest ones.