Affixes in English: analysis and implications
Most EFL researchers will agree that there is a strong relationship between knowledge of word parts and overall vocabulary size, but the research field has been strangely quiet over the last few decades, with only a dozen or so meritable studies being carried out. This is surprising, considering the salience of this issue and its implications. Bauer and Nation (1993) define a word family as consisting of a “base form and all of its derived forms that can be understood by a learner without having to learn each form separately”, e.g. agree, agrees, agreeing, agreement, agreeable, disagrees, disagreement, etc. Affixes, i.e. word parts, are the parts of words that attach to the base word (agree) to create derived forms (agreement, etc). These distinguish between the various members of a word family. In English, prefixes (such as pre- or un-) cause a semantic change in the resulting derived form, whilst suffixes cause a grammatical change.
Inflectional suffixes (such as plural -s, -ing) indicate number and person but do not alter the base word’s part of speech, and derivational suffixes normally cause a change in the part of speech, such as forming the adjective agreeable from the verb agree. If a word contains both a derivational and inflectional suffix, the inflectional will always follow the derivational (e.g. agreements).
There have been many studies attempting to calculate the prevalence of affixes in English, most of which conclude that they are both frequent and useful. White, Power and White (1989) found that by knowing the commonest meaning of the base word, the meaning could be guessed in 60% of words with the prefixes un-, re-, dis-, and in-. When the subjects knew less common meanings of the base word and were aided by context, this rose to 80%. Nagy and Anderson (1984) showed how around 21.9% of all tokens in the corpus of printed school English were inflected, and 12.8% were inflected with a derivational suffix. Bird (1987) studied the origins of the English lexicon, and found that although Germanic words prevail in the first one thousand most frequent (base) words, after that Latin and Hellenic words are more frequent. Words with classical roots are inflected more often, and thus derivation occurs frequently in the English language.
Chronological literature review
Nation (1990: 31) compiled a list of the aspects of knowledge that a speaker (L1 or L2) must possess both receptively and productively to be able to say that they fully know a word. They must master:
1. The spoken form of a word
2. The written form of a word
3. The grammatical behaviour of a word
4. The collocational behaviour of a word
5. How frequent the word is
6. The stylistic register constraints of a word
7. The conceptual meaning of a word
8. The associations a word has with other related words
Later (2001: 27), he amended the list and broke the types of knowledge into categories:
Form
1. Spoken
2. Written
3. Word parts
Meaning
1. The ability to connect the form with the concept
2. Conceptual meaning and referents
3. Paradigmatic relations
Use
1. Grammatical behaviour
2. Collocations
3. Usage constraints, e.g. register, frequency
Based on Nation’s 1990 list, Schmitt and Meara (1997) hypothesised a link between two of these types of knowledge, word associations and grammatical behaviour, and their contribution to overall vocabulary size and general language proficiency. They focussed on a particular aspect of grammatical behaviour; suffix knowledge, as morphological mastery is acquired later than knowledge of word class or sentence order, and thus students capable of using suffixes would have a large enough vocabulary to cope with word associations. They hypothesised that:
“greater knowledge of suffixes would coincide with a larger vocabulary because it would facilitate access to more members of a word’s family. This makes it more likely that an appropriate word family member is available for use as an association (nature might associate with agreeable but not with the more common agree)”
In other words, the more derivatives of a word a learner is introduced to, the larger their vocabulary becomes, and the more their semantic network of word associations in the L2 branches out and expands. They tested 3 groups of Japanese EFL learners, all of whom had received at least 5 years’ instruction in English. From 20 prompt words (all verbs) they were asked to list the allowable affixes from a list of 14 (-ed, -er, -s, -able, -en, -ly, -ion, -ment, -age, -ance/ence, -al, -ee, -ive, and -ure) to test their receptive knowledge. To test their productive knowledge, they were asked to list all allowable affixes for each of the 20 prompt words. The subjects completed the test at the beginning and the end of the academic year. Although some have deemed the improvement of the subjects ‘modest’ (Milton, 2009: ch5), Schmitt and Meara found a statistically significant correlation - 0.41 - between suffix knowledge and vocabulary size. Perhaps unwisely, they were quick to interpret this as a causal relation, positing that suffix knowledge resulted in a growth in vocabulary size. Milton (2009: ch5) points out that this could easily be reversed; a larger vocabulary may be necessary to encounter a suffix enough times and fully appreciate its “significance”.
Mochizuki and Aikawa (1997) were more cautious in their conclusions. They said that Schmitt and Meara’s results in their 1997 study may have been corrupted by the subjects knowing the words rather than the specific suffixes; in other words, Schmitt and Meara tested the subjects’ knowledge of certain verbs and their allowable derivations, rather than actual suffix knowledge. To demonstrate this, they designed a study that singled out suffix knowledge from whole-word knowledge. Based on Mochizuki’s 1998 study, they chose 13 prefixes and 16 suffixes based on Bauer and Nation’s levels 3 to 6 and Nation’s 1996 vocabulary list. For each prefix, they chose three words containing the prefix and made them unrecognisable by changing the consonants. Underneath, 4 different meanings of the prefix were given in Japanese, one of which the subjects had to chose. Here is an example:
redronuce reuzipe refait
(1) together (2) before (3) extra (4) again
For each suffix, three words containing the suffix were made into pseudo words by changing their consonants, and the subjects were asked to select the correct part of speech from four options: noun, verb, adjective, adverb. Here is an example:
rombortable quifable slomitable n. v. a. ad.
They noted an even stronger relationship between affix knowledge and vocabulary size (0.58 with prefixes, 0.54 with suffixes, and 0.65 combined scores). On the whole their test group was much the same as Schmitt and Meara’s (high school and university students, with Japanese as their L1), but the size was much greater (403 compared to 95 in Schmitt and Meara’s).
In a paper submitted to System in October 2008 entitled “Suffix knowledge: acquisition and applications”, (we shall refer to it as the ‘System paper’) researchers conducted a study on an high-ability EFL group (medicine undergraduates) and an medium-ability EFL group (engineering undergraduates), both in a Thai university. The subjects were asked to translate 16 base words and 16 derived forms, from a total of 32 word families. A week later, they were asked to translate 16 derived forms of the base words in test 1, and the base words of the derived forms in test 1. Unlike other studies previously discussed, they used words for the Academic Word List (AWL), as they were focussing on the relevance of affix knowledge in English for Academic Purposes (EAP). It is perhaps more relevant in this aspect than any other, as academic jargon is more formal and is ripe for morphological analysis, as most of the words have classical roots – 85% of the AWL is of Greek or Latin origin.
The System paper found that the high-ability group had a larger vocabulary and showed signs of word building, and the medium-ability group (with a smaller vocabulary) showed almost no word building capabilities. They came to a rather different conclusion from other similar studies; that derived forms – when given - are learnt alongside their base words, but they (i.e. their morphological characteristics) cannot be used to infer the meanings of new words. They advised teachers to focus less on morphological break-down activities and more on the provision of derived forms. They cite Bybee (1995: 430), the emergentist, who said that sets of words sharing semantic and phonological patterns reinforce each other and create emergent generalisations. The G2 learners “clearly” had some mental representations of affixation in English, derived from the frequency of these items. They declared that the word-building strategy in the classroom “may appear to represent a plausible short-cut to vocabulary acquisition”, but is unlikely to be fruitful.
Importance of research into affix acquisition and determinants
Bauer and Nation (1993) explicated the breadth of applications brought by research into word families and affix acquisition in relation to overall vocabulary size. In pedagogy, systematic word-building strategies and a deeper understanding of affixation could increase learners’ vocabularies with less effort than learning each word piecemeal, creating a “graded syllabus” in TEFL. In L2 teaching, when coming across a word in class the teacher can offer some derivatives (e.g. happy – unhappy, happiness), which is much less time-consuming and memory-loading than learning each word individually. Teachers may become attuned to the most useful affixes to teach their students; those that need more emphasis or are more challenging to learn, and which are appropriate to their stage of learning.
This standard definition could also act as a reference point in psycholinguistics and linguistics developmental research, shed more light on the question of lexical storage, viz., is each derived word stored separately in the brain or are derived forms ‘constructed’ to order from their parts?
Bauer and Nation argue for formalising the criteria for vocabulary calculation; in other words, should word families be counted as one word (the base word) or each of its members be counted separately? Some consider any change, including capitalisation, to result in a new word. Others discount inflectional suffixes, others discount both inflectional and derivational suffixes. Standardising the definition of the vocabulary unit and its statistical value in vocabulary size would allow results to be reliably compared. Gardner (2007), however, has some reservations regarding the use of whole word families as a counting unit in vocabulary calculation: there is a risk that a learner’s ability may be underestimated. He says that for a learner from a non-Romance L1, high grammatical competence is required even to master 3rd person singular –s, but this competence would be disregarded in a student evaluation. On the other side of the coin, Bauer and Nation (1993) warn that an overly conservative estimate of word family size would result in many more items to be learned and taught separately.
The main determinants of affix acquisition
Kwon (2005) gives an overview of morpheme studies in the last 4 decades, and discusses driving forces behind morpheme acquisition. She says that the majority of research into determinants was carried out in the 1980’s, but researchers were driven by their own hypotheses and tended to focus on a single determinant, rather than considering a multifaceted contribution to morpheme acquisition. After the historical survey, she foregrounds 3 primary “putative determinants” of morpheme acquisition that have been most frequently cited by researchers are semantic complexity, input complexity, and native-language transfer.
Semantic complexity can refer to affixal polysemy; whether a form expresses more than one meaning. An example would be –th, which occurs in Bauer and Nation’s Level 3 when used in sixth or twelfth, and Level 6 when used in warmth or breadth. Brown (1973) established a complexity hierarchy, in which morphemes with multiple meanings (grammatical or semantic) would be acquired later than those with only one meaning or function. Semantic or conceptual complexity strongly highlights differences between L1 and L2 speakers: L2s already have a concept of time, and so find it easier to learn the past tense, as they only need learn the form and not the concept.
Input frequency denotes the frequency at which a given structure occurs in the
language to which the learner is exposed. There are disagreements over the mechanisms behind this determinant: on one side of the fence are the behaviourists, who say that the learner hears and imitates frequently used structures; and on the other side are the nativists, who say that hearing the structure activates the learner’s Language Acquisition Device (LAD- Chompsky 1981). Larsen-Freeman (1976) propose that input frequency is the principle determinant in the order of L2 acquisition.
The third putative determinant discussed by Kwon is native-language transfer, that similar structures in the L1 will facilitate learning and different structures will deter learning in L2 acquisition. Many studies carried out in the 1970s (including Dulay and Burt 1974, Fathman 1975, and Larsen-Freeman 1976) examined subjects from a variety of L1s in a single study, and harvested results that suggested the order of acquisition does not differ greatly and thus cannot be determined by a learner’s L1. Some studies, however, found elements of L1 transfer in the acquisition of certain ‘foreign’ structures, such as the placement of verbal negation in German EFL learners (Wode 1977), acquisition of English plural -s (Zehler 1982), and the slower article acquisition in a Chinese child than a Spanish child learning English (Zobl 1982). She concludes by saying native-language transfer should no longer be dismissed, that the mechanism “plays an important role in L2 acquisition” (p. 15).