A Study of Receptive and Productive Vocabulary Learning from Word Cards.


ROBERT WARING


Abstract




INTRODUCTION

Much has been written about receptive and productive vocabulary. Crow and Quigley (1985: 499) point out that "most experts acknowledge the importance of distinguishing between active and passive vocabulary". However very little work has been done to distinguish the two. That is, what is different about them, why do we make the distinction? More importantly we must be careful when using the terms 'receptive' and 'productive' and heed Melka Teichroew's warning (1982: 5) that we don't necessarily know what we are talking about when we refer to receptive or productive vocabulary. Nor do people agree what they are exactly. It is not certain by any means that the two can be so neatly separated and indeed it may involve learning along many continuums that seem to overlap enough that knowledge and learning, for convenience sake only, are labelled receptive and productive (Oller, 1979: 424 ff.).

There have been many attempts to measure vocabulary knowledge in terms of scales (Wesche and Paribkaht, 1993) or lists (Gairns and Redman, 1986; Crow, 1986: 243; Richards, 1976), but none of it seems to me, to quite capture the multifarious aspects of word knowledge or what it takes to 'know' a word receptively or productively (see Waring 1996a forthcoming and Waring in preparation). Wesche and Paribkaht, for example, identify 5 categories or levels of knowledge. While 3-5 categories of knowledge may be enough for rough distinctions between types of knowledge for vocabulary testing purposes, they may not be adequate for a more complete description that would be needed if one was searching for a general theory of vocabulary acquisition.

Several attempt have been made to distinguish the two from theoretical and quantitative perspectives. Crow (1986) for example says that the two differ in that productive use requires more accurate mastery of connotations and denotations, register, syntactic constraints, co-occurrence restrictions and so on. Quantitatively, Waring (1996b, forthcoming) looked at the relative vocabulary size of some Japanese learners of English and found a differential between the two. He found that if a high frequency word was known receptively, there was good chance (64%) that it would be known productively. However, if a low frequency word was known receptively there was little chance (15%) it would be known productively. This means that we know far many more words receptively than productively, but that there is no linear relationship between the amounts known. Similarly, Eringa (1974) and Holden (1890) looked at productive vocabulary size, while others tested aspects of receptive vocabulary such as Groot and Hoekstra (1981); Seashore and Eckerson (1940); Meara and Jones (1990); Hartmann (1946).

The above tends to support the commonly held notion that productive word knowledge is more elusive, more difficult to learn, and possibly more fragile. There are several ways to quantify this. One could look at the relative speed of learning in terms of time taken to learn a given number of words and then at the retention and relative decay. From the above, Hypothesis 1 was derived .


Hypothesis 1: Productive learning of a given number of items will take longer
than receptive learning.

This tries to answer the question 'which takes longer, receptive or productive learning?'. The longer the time it takes will give an indication as to which is more difficult to learn. This hypothesis can be tested by asking learners to learn 2 sets of words, some words receptively and some productively and then comparing the time taken to do so. It is assumed that a learner has more to do when learning a word productively and it will take longer to learn a given number of words productively than receptively.

A complementary question is, 'do learners remember more about the receptive or productive aspects of words?' If we assume that one has to do more to learn a word productively, tests of the learning will yield more words recalled if tested receptively than if tested productively. That is receptive test scores will be higher than productive test scores.

Hypothesis 2: Tests of word learning reveal higher scores for receptive knowledge than for productive tests.

Carter and McCarthy (1988: 94) point out that although there is a useful distinction between receptive and productive vocabulary, the two may not be automatic polarities. It is commonly held that comprehension precedes production (see Channell, 1988 for just one example) Clark (1993: 246) says that 'logically, comprehension must precede production'. Here she equates comprehension with an understanding of the word's meaning, and it seems logical to assume that one must need to know what it means to use it. However this observation is by no means absolute. Carter and McCarthy also raise the question of whether production can indeed precede comprehension (as did Teichroew, 1982). It may be that a word can be used without fully comprehending its different meanings say by mistake, or as a result of incomplete comprehension. Testing out and creatively playing with words and their meanings in the process of learning, what Clark (1993: 247) calls the 'alignment process' between new data and the linguistic community, could also mean that sometimes production precedes comprehension and I feel safe in surmising that most language learners have experienced this. This testing of productive knowledge can only be done contextually when a learner is experimenting with his knowledge in context and with an interlocutor. It is somewhat less clear whether comprehension precedes production for the decontextualized learning of words.

Some researchers state that as one learns a word, the receptive knowledge comes before the productive (Ingram, 1974, Bloom, Clark, Hutcheson and Van Burrel, 1974; Ringbom, 1985) As is common in many articles, the terms are not defined and the lack of a definition of these terms leaves us without a specific area to test. This aside, it may be that some words can be learned receptively (by learning the L1 equivalent of a L2 word) but not be recalled on a test of that knowledge, but at the same time she can retrieve the word from the L1 prompt and write it correctly. Here, production would precede reception. One may recognize a word without recalling the meaning, or one may recall the meaning in one context, but not even recognize it in another. 'Reception' seems to be a moving target unless defined explicitly, illustrating Melka Teichreow's point mentioned above.

The 'receptive before the productive' position is put forward by Ringbom (1985: 168) where he says "items pass from the learner's receptive vocabulary store to his smaller productive one all the time, some items perhaps moving back again from the productive to the receptive one, when the learner forget items he has once mastered but has not met very frequently or recently". This assumes that there are two stores and that the two are separate and that loss of productive control means a retreat ('moving back') to the receptive store where receptive control is still retained. This therefore assumes that productive word knowledge encompasses the receptive as a subset. This means that if a word had been learned productively it would logically be known receptively as well. Simplistically, this serial view of receptive and productive competence would look something like the following.


Figure 1: A serial view of levels of receptive and productive word knowledge.

Word Meaning Retrieval Ability Ability to
Recognition Recall of the word to spell or use the word
from memory say the word appropriately
in context


------------------------------------>
RECEPTIVE COMPETENCE PRODUCTIVE COMPETENCE

It is my contention that receptive and productive knowledge is not serial in nature or indeed organized into subsets or different lexicons, but is actually a far more complex system involving associations, and having multi-dimensional aspects of language knowledge and is at least in part, a network-like phenomenon (see Waring, 1996a, forthcoming, Waring, 1994). This view sees a series of overlapping degrees and levels of knowledge continua and levels of automaticity that would allow receptively learned words to be produced and also it would allow productively learned words to not be known receptively (not be available for recognition or recall). For example, it is entirely possible that a word may be available for retrieval for speech but may not be recognized or recalled in reading, or that a word is known in one context but not another. Admittedly, these occurrences would be smaller in number than words known receptively but not productively, but nevertheless they exist and any general theory of second language vocabulary acquisition needs to account for this.

If receptive and productive knowledge are separate and distinct as defined above, then learning a word 'receptively' would require work on recognition and recall but aspects of productive competence such as 'retrieval' and 'use' would not need attending to as there is no requirement to do so. If it is found that a word learned receptively (from an L2 prompt and the meaning was correctly recognized and recalled - say by writing the L1 equivalent) can also be produced when asked to produce the L2 word from an L1 prompt, then there will be a flaw in the above serial or subset view. This is because the L2 to L1 learning had left a memory trace sufficiently strong enough to allow L1 to L2 production, albeit in a limited way. This can be tested by asking learners to learn some L2 words receptively, but test them productively. It is my assertion that some words learned 'receptively' are available 'productively'. The converse is that not all words supplied 'productively' will supplied (known) 'receptively'.

Hypothesis 3: Some words learned receptively (L2 to L1) will also be available
productively.

Hypothesis 4: Some words learned productively (L1 to L2 learning) will be able
to be recognized and their meaning recalled when tested receptively (L1 recognition from an L2 prompt).

This can be measured by asking learners to learn some L2 words productively, but test them receptively. If a word can be understood (for example, if an L1 translation is provided that demonstrates an understanding of the L2 word) when it was learned productively, it shows that learning words 'productively' does not restrict the learner to productive use only. In Hypotheses 3 and 4 it was stated that 'some words' will be learned not 'all words'. This is because clearly, there can be no definitive learning that occurs with all words on all occasions.

If we follow Crow's (1986) reasoning that more about a word has to be learned to learn a word productively, it would take more time to learn a word productively than receptively. But at the same time all this extra information needed to be learned could render the learning more fragile than receptive learning as mental resources could be stretched thinner due to the volume of knowledge needed to be retained. Soon after learning it would be expected that quite a high percentage of productive knowledge would be retained, with it decaying after some time. It would also be expected that as time passes, more productive word knowledge would be lost than the receptive.

Hypothesis 5: Information used for the production of a word will be forgotten more quickly than information needed for recall.

This tries to seek an answer to the question, 'which is more fragile, the receptive or productive learning?'



THE STUDY

Intention
This study was undertaken to look at the learning of word pairs in a decontextualized setting. Learning from word cards is a commonly used and linguistically appropriate strategy for the fast encoding of new vocabulary (Mondria and Mondria De Vries, 1994). The intention of the study was to measure the speed of learning from word cards and the retention thereof both receptively and productively. The difficulty of learning and the shape of what is learned and retained is also investigated.

Definition of receptive and productive learning.
For the purposes of this study of decontextualized learning, to demonstrate 'receptive knowledge' of a word, the subject must be able to provide a specific L1 translation of the English word which shows an ability to recognize the word and recall the meaning learned (L2 to L1). To demonstrate 'productive knowledge' the subject must provide a specific L2 equivalent for the L1 word showing an ability to recall the equivalent L2 word and have control over its spelling (L1 to L2). Therefore for the purposes of this study, 'receptive' knowledge refers to recognition and recall and 'productive' refers to retrieval only.

Method
The subjects
70 female 18 to 20 year old Japanese learners of English at private university in Western Japan were used as subjects in this study. All learners had studied English for about 6 or more years and were studying English as their major at the university. The 70 subjects comprised 2 intact classes at the university. The diversity of the students was not a factor as the study was a within subject design. All subjects were familiar with word card learning, mnemonic techniques and so on as vocabulary learning strategies were one of the core components of the class. Regular sessions were being given within class time where the subjects could exchange ideas, hints and mnemonics. Due to absences for post tests the data of only 63 subjects were collected in full, but unfortunately they did not comprise a balanced statistical design so the data for three subjects were randomly dropped to make the design balanced.

The instrument
The subjects were required to learn 30 English-Japanese word pairs that were chosen by the researcher (details below). 15 of the words were to be learned from Japanese to English (the productively learned words) and 15 different words were learned from English to Japanese (the receptively learned words). Words were presented in two sets of 15 words (set A and set B) that were to be learned separately. The word pairings were on word cards (also known as flashcards) - one card per pair - with one side having the Japanese word and the other the English. All the cards were carefully arranged so as to appear the right way up and the right way round. Each set had a 'learn from English to Japanese. Start here' (or vice versa) label on them. This was so that the learner would be learning all the words starting with the English or the Japanese word appearing first as they progressed through the set.

The words
The words were not randomly selected but were carefully chosen for many reasons. The words chosen to be learned had to be known in Japanese but not known in English so as to show no effects for prior knowledge or lack of it. From previous tests that the classes had taken it was determined that words taken from the Thorndike and Lorge (1944) 'less than 4 times in a million' category should be used which corresponds to a vocabulary frequency word far in advance of these learners. A pretest of these words was administered as detailed below.

Each set of 15 words was made up of 6 2-syllable words (3 with final and 3 with initial stress), and 9 3-syllable words (3 each with final, middle and initial stress). 5 nouns, 5 verbs and 5 adjectives made up each set. Although some words belonged to more then one part of speech, the translations given made up equal numbers of the nouns, adjectives and verbs in Japanese. A worksheet was made, copied and cut up to make word sets. The full set of words appears in the appendix.

Sets and order.
To offset the possibility of effects for the order of learning (either receptive first or productive first) the sets were alternated with half the subjects having to learn receptively first and the other half productively first. Similarly, to offset the possibility of effects for the set being learned first (either set A or set B) the sets were also alternated with half the subjects having to learn set A first and the other half set B first. For data analysis purposes only this made groups of 4 subjects - the first person learned set A receptively first, the second person learned set A productively first, the third person set B receptively first and the fourth person set B productively first.

Procedure

Piloting.
The number of words to be learned was chosen to be 15 due to time constraints. The experiment was conducted in a class time of 90 minutes. In that time there needed to be a pretest, instruction about the experiment, demonstration time, two sets of words had to be learned needing unlimited time to complete, a distractor and 2 post tests again allowing as much time as possible. It was assessed and confirmed in the piloting that 90 minutes would be needed by the slowest of the learners.

The English words were chosen with the help of 8 students in the same academic year (but not in the 2 classes) to see if the English target words may be known by their peers. At the same time it was also checked that the 30 translations were all known in Japanese by them. Several revisions had to be made before a final set was chosen of which they knew none of the English words but they knew all the Japanese words. The final instruments were tested using the procedure outline below with 4 subjects. Minor alterations to instructions were made as a result. No other changes were found to be needed as a result of the piloting.

Pre-test Introduction.
The subjects were not informed that they would be tested for knowledge of the target words first and were given the pre-test (see below) as a surprise 'test'. These 'tests' were then collected and kept and make up the baseline data. The subjects were then told about receptive and productive learning and the upcoming experiment. The experiment was conducted with the intention to help them find out how quickly they learn words, which way they learned better and faster. After the delayed post tests they also learned how quickly they forgot them and became more aware of the need for spaced repetition in their learning to help retention. All students were given the data for each of their tests to keep after the fourth post test.

Pre-test.
As mentioned above, a list of the 30 words in English was given to the learners as a pre-test to see how much was known. The subject was required to write an approximate Japanese translation for each word, or if unable to do so was required to underline the word if she 'had seen it before but can't remember the meaning'. It was decided not to conduct a productive pretest because the Japanese words given may have elicited words which were not the target words. This is because a word such as 'flunk' could be translated as the more common 'fail', or 'duff' as 'broken' and it would not be clear if the less common words would be known.

The experimental procedure.
Each subject was given an envelope containing 2 smaller envelopes one labeled 'first' and the other 'second'. One envelope contained set A and the other set B. As mentioned above, each set of words had a label giving instructions which way (Japanese to English or English to Japanese) they were to be learned. The envelopes were distributed amongst the subjects in such a way that allowed the researcher to observe which direction the words were being learned. During the experiment the researcher did not find anyone who was learning in the 'wrong direction'. The subjects were told that the criterion for learning was met when the learner had gone through the 15 words in each set 2 times without mistakes. The start and finish time for both sets was recorded by each subject. Unlimited time was given for the subjects to complete their learning. The subjects were instructed not to write during the experiment and thus had no list of words to take with them to affect the delayed post tests. For homework, the subjects were required to write a short essay about their learning; what they did during the experiment; how they learned; what they found easy or difficult to do; and what they had learned about their learning style and so on.

Post tests.
Between the experiment and post test 1 (which was on the same day of the learning), a 10 minute silent comedy movie was shown as a distractor. The post tests were all a surprise - no forewarning had been given. There were two parts to each post test - a receptive one and a productive one. The receptive post test was the same as the pre-test - a list of the English words which needed to be translated into Japanese (a translation would show a receptive understanding of the English word). The productive test was a list of the Japanese words for which the English translation was required (showing the ability to produce the English word from the Japanese translation). On each occasion the productive test was given first in order that the subjects would not see the English words on the receptive test and be reminded of them. The subjects were told to write the word they had learned during the experiment (not a close translation which may have been acceptable). Unlimited time was given for the completion of all the tests. Delayed post tests were given on the following day, one week later, and three months later making 4 post tests. However, for post tests 2, 3 and 4 the order of the words was changed from post test 1 on the answer sheets to offset the possibility that the subjects may remember which word goes where on the answer sheet from the day before. They were made aware that the positions of the words on the page had been altered.

Scoring
For each subject, 9 worksheets had to be marked - a pretest and 4 productive and 4 receptive post tests. The pretests and the productive tests (requiring the production of English words) were marked by the researcher - an English native speaker. The receptive test (requiring the production of Japanese words) were marked by the researcher - a native speaker of Japanese (who had helped compile the list of Japanese word pairings and was familiar with the study). Both sets were checked by one other person and in times of disagreement joint decisions were made.

It was the researcher's intention that as the tests were rather insensitive to small amounts of learner knowledge, the scoring should be sensitive to small amounts of learning. A full one point was given for a word given as it appeared on the original word card. No point was given for a blank or incorrect answer. Half a point was awarded for spelling or Kanji mistakes where it was clear that the word was at least partly known. Where words were misspelt by more than 2 letters no points were given. Lists of words that were awarded half points were kept to refer to during marking to maintain consistency.

DATA ANALYSIS

The data relating to the above hypotheses will be presented in this section, discussion however will be left until later.

Pretest results
Most subjects handed back blank sheets in the pretest. Some words were translated but most were incorrect guesses or interferences from similar sounding words. Of the 30 words attempted by 70 subjects, only 3 words were correctly translated into Japanese, one each by 3 different subjects (their data were not used in the final analysis presented here). A few (5-6) subjects underlined 2 or 3 words with one person underlining 8 (she was absent for 2 of the post tests so here data were not used in the final analysis). Interviews with these subjects revealed that in fact they had not known the words they underlined when asked to either explain it or produce a sentence with the word. It can be safely stated that the 30 words were unknown prior to the learning by the 60 subjects.

Rate of learning by word cards
There was considerable variation in terms of how much time was taken to learn these words. The fastest learner took 630 seconds to learn the 30 words (15 receptively and 15 productively) at a rate of 171 per hour (not allowing for forgetting and fatigue over a full hour - however see studies by Thorndike (1908); Webb (1962) and Anderson and Jordan (1928) who showed that decontextualized word-pair learning rates increase with the time spent doing it. By contrast, the slowest took 2321 seconds to learn the same 30 words (46 words per hour). Figures 1 and 2 show the distribution for the subjects.
@

FIGURES 1 and 2
@


A Spearman rank correlation of 0.29 was found for the correlation between the receptive and productive learning times. The times taken to learn set A and B were compared to check for effects due to the set and order learned. A 2 (set A or B) by 2 (receptive first or productive first) by 2 (time taken to learn receptively and productively) repeated measures multivariate analysis of variance (MANOVA) was calculated with sets and order as within subject variables and time taken as a between groups variable. No effect for set F= 0.02, (1, 56) p 0.96 or order F = 0.28, (1, 56) p 0.60 was found. A significant main effect for the time taken for productive learning was found F = 113.2, (1, 56), p 0.0001, meaning that productive learning took longer. The means are presented in Table 1.

TABLE 1 ABOUT HERE

For all the 60 subjects, learning words productively (717.0 seconds) took 26.1% longer than doing it receptively (568.6) irrespective of which set was learned or the order in which they were learned. This supports hypotheses 1 and 2 by demonstrating that a learner has more to do when learning words productively.

Data were then analyzed for the learning by looking at the post tests to see whether learners remember more about the recall or retrieval words. The descriptive statistics are in Table 2.

TABLE 2 ABOUT HERE


A 2 (set A vs set B) X 2 (order 1 vs order 2) X 2 (type of learning) multiple analysis of variance (MANOVA) was administered on the data with set and order as between subject variables and type of learning as within subject variable. A significant main effect for type of learning with receptive learning yielding higher test scores F = (1, 56) 115.2 p .0001. No significant effect was found for order F = (1, 56) .05 p = 0.89 or for set F = (1, 56) .00 p = 0.95.

From the above, the order and set variables were not significant. From these data we see that there is little difference in scores on any post test between sets A and B whether learned productively or receptively, which indicates no effect for the set learned in terms of retention. It is clear that receptive learning leads to higher scores or that productive learning leads to lower scores, supporting Hypotheses 1 and 2.

The data were re-analyzed by looking at the type of learning and the learning outcome. Table 3 shows the results of productively learned words tested productively and tested receptively as well as receptively learned words tested both ways. This was done to see whether there is a difference for learning type and testing type. It tries to answer the question 'are productively learned words better remembered if tested productively or receptively, and are receptively learned words better remembered if tested productively or receptively?' It also seeks to quantify the percentage of words known by type of learning.

@

FIGURE 3



Table 3 shows that for post test 1, 32% (6.07/ 12.83 + 6.07) of the words recalled were not learned with the intention of recalling it. Similar percentage over the remaining post tests. To put it another way, 40.4% (6.07 /15) of words learned for recall are available for retrieval, with this knowledge decaying slower than productive words tested productively. These data are shown graphically in Figure 3. One must be careful to notice that Figure 3 is distorted graphically in that the first 4 items all occur within a week, but the last one 12 weeks later showing a significant drop. We must also be careful not to interpret that there was a huge drop in productive knowledge for the receptively learned words as the subjects did not have to learn these words productively and therefore no 'consciously learned knowledge' was lost. However, there seems to be some transfer of knowledge incidentally which allowed for productive knowledge to be shown for receptively learned words, supporting Hypotheses 3 and 4.



DISCUSSION

In this section I will discuss and interpret the data from the experiment. First I shall discuss the benefits of word card learning per se before moving on to a discussion of the comparison of the relative ease or difficulty of each type of learning.

Benefits of word card learning.
The time taken to learn the 30 words averaged 21 minutes 28 seconds (566.5 + 721.4 = 1287.9 seconds) with half learned receptively and half productively. This translates into 84 words per hour for decontextualized learning assuming no effect for fatigue (60 / 21 minutes 28 seconds = 2.8 X 30 words = 84 words per hour). It has been seen that in fact there is an advantage for the time taken to learn decontextually. Receptively this equates to 95.3 words per hour but productively 74.8.

The retention over three months for one hour of word card learning, tested decontextually would be about 49 words or about 60% (total of mean scores for both tests after one week from Table 3 divided by 2 (18.15) as a percentage of the maximum score (30 words) multiplied by 2.8). This is without any further practice in the meantime. Although these figures are somewhat rough, clearly a serious program of word card learning can result in huge vocabulary gains at least at an initial stage of word learning. The question remains however as to how much of this decontextualized initial learning can be transferred to contextualized language tasks and real life use, this awaits a further study.

Receptive learning is easier than productive learning.

The data in tables 1 and 2 show there is a distinct advantage for receptive learning with it yielding a 36.6% better suppliance than the productive on the first post test (23.65 / 17.31) as per Table 2. This is despite the receptive learning taking 26% longer as outlined in Table 1. This difference grows with the length of time from the learning and by the 4th post test the advantage had increased to 515% (13.4 / 2.59). Both these sets of data support hypotheses 1 and 2, clearly productive learning takes more time and is less well retained.



FIGURE 4



The rank correlation of 0.29 for the interaction between receptive and productive learning times indicates little support for the view that learners are equally capable of learning receptively and productively. If they were, then each learner would be at the same rank for both tests. That is some learners can learn one set quickly while the other takes more time. This is shown by the data in Figure 4. The data show large variation within individuals. Some subjects show very close scores while others are vastly different ones.

Why would this be so? Some of the attempts to distinguish receptive and productive vocabulary knowledge state that the cognitive load of learning a word receptively is higher than that for the receptive learning. This is because when one learns something productively one must learn more things including the spelling or pronunciation. In contextualized learning one would also have to be able to use the word contextually, both pragmatic and linguistic (see Crow, 1986 above).

Serial or subset view, or something more complex?

The notion that receptive knowledge precedes the productive has an intuitive appeal but I feel misses the mark. The logical extension of this is that one cannot do anything productively until a word is known receptively, and if a word is known only receptively it is not available productively so in effect receptive knowledge is a subset of the productive. However, it may be possible to use a word productively say in speech, without being able to recognize it or recall it receptively, say in reading. Moreover it may be possible to retrieve it for use on a written test but not recall its meaning when met in reading. This is exactly what was found here.

Under the serial or subset view of vocabulary acquisition as outlined in the introduction, receptive knowledge should not be available productively and zero scores on productive tests would be found as this knowledge would not be available for production as it is outside the subset. However, this is not the case as over the first week approximately 6 of the 15 words learned receptively were able to be correctly produced in written form on the post tests. Following a subset view that receptive knowledge is part of the productive, high scores for productively words tests receptively were given (approximately 11 of the 15). Interesting is the difference between these two. Higher scores were able to be given for productive learning (11.08 + 10.76 = 21.84) compared with having learned the words receptively (6.05 + 12.70 = 18.75) a 16.5% difference despite productive learning taking 27% longer to do.

Forgetting / Decay

Overall, the figures for decay closely resemble those of other studies of forgetting. I shall just mention two here. Anderson and Jordan (1928) who only measured recall, not retrieval, over 4 spaced post tests found that knowledge decays at a relatively even rate after an initial loss. They assessed learning immediately after learning, after one week, three weeks and eight weeks and found that the percentages retained were 66%, 48%, 39% and 37%. Although the intervals here are different from the present study, the basic shape remains in that it does not match Ebbinghaus's (1885) famous forgetting curve that predicted far larger losses soon after learning (up to 70% by 24 hours). From studies of the rate of forgetting of languages and words (see Bahrick, Bahrick and Wittlinger, 1975; or Bahrick, 1984 for just 2 examples) we could assume that after three months, recall rates would be 20% or less of the initial learning, however the results found here were substantially higher. In fact, it supported Bahrick's (1984) findings that there were substantial amounts of knowledge retained (at least for the receptive knowledge) over a long time. This led Bahrick to suggest that some words are retained in a permastore (stemming from an analogy with permafrost which is unchanging in nature and very durable). It seems then that Bahrick's notion of a permastore of words may have a ring of truth to it.

When one compares receptive against productive knowledge decay, receptive word knowledge showed more resilience than productive word knowledge. It declining less over the three months (46% and 39% to 87% and 81% - as measured from the 'same day' test) by 3 months, which supports Hypothesis 5 that productive knowledge is less resilient. However, the productive knowledge tested productively declined faster over the first week and had almost disappeared by three months. An analysis of the data reveals that 3 exceptional subjects who scored 6, 5, and 4 on the 4th productive post test distorted the means for that test. In fact, 60% of the subjects showed little or no knowledge at all after 3 months for this category scoring 0.5 or less. This would lead to the conclusion that repetition or rehearsal is more important for retaining decontextualized productive knowledge, than for receptive.

Variation

Another factor that hugely affected the data was the huge variation in subject performance both between and within subjects as evidenced for the data. This is rather worrisome. This means that one cannot reliably state what has happened unless one looks at an individual level. There is a need to analyze studies like this from 2 perspectives. The first is to see how the suppliance of words changed over the 4 post tests. Were the same words supplied? Did some appear and then disappear? What exactly is going on here? Secondly, one needs to analyze the data from these studies in terms of the total percentages of words supplied receptively only, productively only and both ways across the 4 post tests. This will allow us to see the distribution of the learning and how it changed.

PROBLEMS

Several cautions must be made when interpreting this study. As mentioned above, this was an exploratory study looking at the decontextualized learning of word pairs from L1 to L2 and L2 to L1. As such the intention was to identify areas for future research and clarification. The first problem is that the study looked only at a very small number of words. Nagy and Herman (1987) and echoed by Meara (1996: 38), - are quite right to say that the testing of the learning of 30 as a subset of tens of thousands in English does not validate a particular study nor lead to the formation of a cornerstone of data in the field.

Secondly, the test design may have left a memory trace of which words were to be recalled on the receptive one from the productive test. The subjects did the productive test first on all post tests and did not see a list of the L2 words until they were presented on the receptive test. However, the list of Japanese words formed a set of words to be used on the corresponding receptive test, this memory trace may have prompted recall of the set. A subject may have said when looking at the receptive test containing the English words, 'Oh, that's right! Now I can recall which Japanese word goes with which English one, because it must be one from the other test'. Without the L1 list, it may not have prompted the recall of which Japanese word needed remembering, triggering a better than normal recall rate. This could have been averted if the productive test had been given some time apart from the receptive, say 2 or 3 days apart. Unfortunately this was not possible. This is not a serious problem as the words were for recall. This was also the reason no recognition test was administered (for example 'underline all the words you have seen before') because if it had been, the full list of L1 words on the productive test would be immediately available for recognition and it would have been a simple task just to underline the entire list.

Thirdly, the learning and testing was done decontextually. This does not invalidate the results found in this study because the kind of learning done here is a typical strategy used by many second language vocabulary learners. However, the results do not allow a transfer of these findings to vocabulary learning per se as so many of the words we learn are learned unintentionally and from context. Further studies need to be done that measure receptive and productive learning in context.


CONCLUSION
Some directions for future research have been outlined above. It seems that the major finding of this study is not the nature of receptive or productive vocabulary itself, but a discovery of enormous variation within and among individuals in respect to these phenomena. Without a clear picture of the reason for this, and some indications of what we can do about it, studies in this area will languish from lack of applicability to real language learners and classrooms. We must understand the learners and in a systemmatically and principled way, before we attempt apply our theories to them. This it seems to me, is the next step.

The results here have indicated that there is no neat divide between so called 'receptive' and 'productive' learning. In fact knowledge from receptively learned words can be available productively and vice versa. It seems then that a complete description of receptive and productive vocabulary knowledge must account for this. We must however heed Meara's (1996) warnings that to look into these matters without a principled overall aim compatible to most of us working in the field of second language vocabulary acquisition, we risk further fragmentation of the field "making it difficult for us all to pull all our findings together in one coherent picture" (p. 38).

APPENDIX

A list of the words used on the test (note some words can be two or more parts of speech and blustery can be said with 2 or 3 syllables).

flunk kip duff guzzling shambles skimpy
bemuse caboose irate amortize clodhopper decadent
embezzle fiasco besotted scram ghoul swank
hanker catsup prudish recoup cartel sure-fire
allocate sturdiness blustery cremated compulsion bombastic


REFERENCES

Anderson, J. P. and A. Jordan. Learning and retention of Latin words and phrases. Journal of Educational Psychology. 19, 485-496. 1928.

Bahrick, H. P. Semantic memory content in permastore: fifty years of memory for Spanish learned in school. Journal of Experimental Psychology: General. 113 (1): 1-37. 1984.

Bahrick, H., P. Bahrick and R. Wittlinger. Fifty years of memory for names and faces: A cross sectional approach. Journal of Experimental Psychology: General 104: 54-75. 1975.

Carter, R. and M. McCarthy, (Eds.). Vocabulary and Language Teaching. Longman, London. 1988.

Channell, J. Psycholinguistic considerations in the study of L2 vocabulary acquisition. In: R. Carter and M. McCarthy (Eds.) Vocabulary and foreign language learning. London: Longman. 1988.

Clark, E. The Lexicon in Acquisition . Cambridge Studies in Linguistics, 65. Cambridge: Cambridge University Press. 1993.

Crow, J. Receptive Vocabulary Acquisition for Reading Comprehension. Modern Language Journal. 70, 3: 242 - 250. 1986.

Crow, J. T. and J. R. Quigley. . A semantic field approach to passive vocabulary acquisition for reading comprehension. TESOL Quarterly, 19, 3: 497-513. 1985.

Ebbinghaus, H. Über das Gedächtnis. Leipzig: Dunker. (Translation by H. Ruyer and C. Bussenius, (1913), Memory. New York: Teacher's College, Columbia University. 1885.

Eringa, D. Enseigner, c'est choisir: vocabulaire-verwering. Levende Talen. 306: 260 267. 1974.

Gairns, R. and Redman, S. Working with Words. Cambridge University Press. 1986.

Groot, P. M. J. and J. G. Hoekstra. . Tests of English vocabulary command for EFL students at University level. Toegepaste Taalwetenschap in Artikelen. 11: 98-136. 1981.

Hartmann, G. W. 1941. A critique of the common method of estimating vocabulary size, together with some data on the absolute word knowledge of educated adults. Journal of Educational Psychology. 32: 351-358.

Holden, E. S. On the number of words used in speaking and writing. Bulletin of the Philosophical society of Washington, II, (Appendix V1). p. 16. 1890.

Meara, P. and G. Jones. 1990. The Eurocentres Vocabulary Size Test. Zurich: Eurocentres.

Meara, P. The Classical Research in L2 Vocabulary Acquisition. In G. Anderman and M. Rogers. Words words words. Multilingual Matters: Cleveland. 1996.

Melka Teichroew, F. J. Receptive versus productive vocabulary: a survey. Interlanguage Studies Bulletin. 6, 2, 5-33. 1982

Mondria, J. and S. Mondria-De-Vries. Efficiently Memorizing Words with the Help of Word Cards and Hand Computer: Theory and Applications. System. 22 (1): 47-57.1994

Nagy, W. E. and Herman, P. A. Breadth and depth of vocabulary knowledge: implications for acquisition and instruction. In McKeown, M. G. and M. E. Curtis. (Eds.). The Nature of Vocabulary Acquisition. Lawrence Erlbaum Associates. Hillsdale: New Jersey. 19-35. 1987

Richards, J. C. The role of vocabulary teaching. TESOL Quarterly. 10, 77 89. 1976.

Ringböm, H. On the distinctions of item learning vs system learning, and receptive competence vs productive competence in relation to the role of L1 in foreign language learning. In; H. Ringböm (Ed.) Psycholinguistics and Foreign Language Learning. Åbo: Åbo Akademi. 1985.

Seashore, R. and L. Eckerson. The measurement of individual differences in general English vocabularies. Journal of Educational Psychology. 31, 14-38. 1940.

Thorndike, E. L. and I. Lorge. Teachers College, Columbia University. The Teacher's Word Book of 30,000 Words. 1944.

Thorndike, E. L. Memory for paired associates. Psychological Review. 15, 122-138. 1908.

Waring, R. A comparison of the Receptive and Productive Vocabulary Sizes of some Second Language Learners. Immaculata, 1. 1996b, forthcoming.

Waring, R. A taxonomy of receptive and productive word knowledge. In preparation.

Waring, R. Connectionism and Second Language Vocabulary. Temple University Occasional Publications. 1996a, forthcoming.

Webb, W. B. The effects of prolonged learning on learning. Journal of Verbal Learning and Verbal Behavior. 1, 173-182. 1962.

Wesche, M. and T. Paribakht. Assessing vocabulary knowledge: depth versus breadth. 1993 forthcoming.




Figure 1: Numbers of subjects by time Figure 2: Numbers of subjects by time
taken to learn words receptively. taken to learn words productively.


o
Table 1. Seconds taken to learn words productively or receptively by set and by order. (standard deviations are in parenthesis).

Set
Set A Set B Total
------------------------------------------ ----------------------------
Productive Rec first 672.3 (247.1) 816.7 (288.6) 744.5 (274.0)
learning Prod first 697.0 (254.5) 699.6 (338.3) 698.3 (294.1)
------------
Sub total 684.6 (246.8) 758.1 (314.7) 717.0 (282.0)

Receptive Rec first 586.7 (186.6) 577.5 (154.6) 582.1 (168.4)
learning Prod first 525.4 (173.9) 576.4 (150.8) 550.9 (160.2)
------------
Sub total 556.0 (179.9) 577.0 (150.1) 568.6 (168.6)

Table 2: Mean scores for Set A and Set B for the 4 post tests. (Standard deviations are in parenthesis.)

Same day One day delay One week delay 3 month delay
post test post test posttest post test

Set A Productive 8.44 (4.4) 7.93 (3.8) 6.08 (3.9) 1.11 (1.3)
Set B Productive 8.87 (4.3) 8.20 (3.8) 7.05 (3.5) 1.47 (1.3)---------------------------------------------------------------
-------------------------------------------
Overall Productive 17.31 (6.0) 16.13 (6.0) 13.14 (6.7) 2.59 (2.3)

Set A Receptive 11.79 (2.8) 11.79 (2.6) 10.99 (3.3) 6.02 (2.9)
Set B Receptive 11.86 (2.5) 12.00 (2.2) 11.40 (2.8) 7.38 (3.2)---------------------------------------------------------------
-------------------------------------------
Overall Receptive 23.65 (4.5) 23.78 (4.5) 22.40 (5.7) 13.4 (5.8)

Table 3: Post test performance by learning type (Maximum 15) and decay over the three months. Data for percentage decay reflect decay from the 'same day' test. (Standard deviations are in parenthesis.)

Testing method Percentage Average
----------------------------------------------- --------------- of decay decay
Same day One day One week 3 month over the per
delay delay delay three months week
---------- ------- ---------- ----- --- --------------- ------------
Receptive learning 12.83 12.21 11.43 6.89 46% 3.8%
tested receptively (2.2) (2.2) (3.1) (3.3)

Productive learning 10.82 11.56 10.96 6.52 39% 3.3%
tested receptively (2.7) (2.6) (3.0) (3.0)

Productive learning 11.20 9.91 7.84 1.41 87% 7.3%
tested productively (3.3) (3.4) (3.7) (1.4)

Receptive learning 6.07 6.22 5.29 1.18 81% 6.8%
tested productively (3.4) (3.2) (3.4) (1.3)



Figure 3: The decay of word knowledge by learning type and test type.


Figure 4: The relative difference between receptive and productive learning times by subject. (Receptive times are ranked, productive times are those relative to the receptive.)

Contact Info:
Rob Waring
Notre Dame Seishin University, 2-16-9 Ifuku-cho, Okayama, Japan 700
Tel 086 252 1155 Fax 255 7663 Home 086 223 0341
Email:Rob Waring


Return to Main menu of papers