Monday 23 April 2012

Which way round?

Rich and famous or famous and rich?

Why do we say rich and famous rather than famous and rich? Or profit and loss rather than loss and profit? Born and bred and past and present, but not bred and born or present and past?

Several reasons have been suggested, but they have never been put to the test. Sandra Mollin has now done this, by looking at the most frequent word pairs that occur in the 100 million word British National Corpus. She analysed 544 different word pairs, which together occurred more than 85,000 times in the corpus.

Mollin found that the most important factor determining the order of words in the pair was to do with their meaning – a semantic constraint. Three different kinds of meaning relationships were relevant. Iconicity explained the ordering in the pair best: this is when the order of the words follows a chronological sequence or a cause and effect sequence, as in spring and summer or born and bred. Power was also important: words referring to more powerful concepts are more likely to come first in the pair, as (so, men and women) and so are concepts that are more central in our society (food and drink, then, rather than drink and food). Perceptual markedness was also relevant: the order of words in the pair often follows what humans seems to perceive as a natural order, such as seeing positive as more important than negative (good and bad), or seeing things that are higher up as more important (head and shoulders, top and bottom). Mollin found that clashes between semantic factors were very rare and in fact only involved gender pairs such as mother and father or bride and groom. In these cases the male should be first in the pair from the perspective of power, but the female is typically perceived as more central in the parental role than the male, and is the more prominent person in a wedding, so it is perceptual markedness that wins out in these cases.

Would you like to help Sandra Mollin with her research into word pairs? If so, there is a survey at
that should take about only 6 minutes of your time.

The next most important factor was a metrical one, to do with the rhythm of speech. This means, for example, that the longer of the two words will come last (so, out and about or past and present), and final stressed syllables are avoided (gas and electricity rather than electricity and gas). When metrical ordering factors clash with semantic factors, it is semantic factors that usually decide which word comes first. Often, though, metrical and semantic factors coincide.

The frequency with which individual words are used is also relevant. More frequent words tend to come first in the pair (so, up and down, or time and effort).

If none of these factors can account for the ordering, phonology (sound structure) comes into play. For example, if one of the words has more consonants at the beginning, this word comes second (so, ups and downs, where up begins with a vowel so has no initial consonant at all whereas downs has one initial consonant; or coal and steel, where coal has one consonant at the beginning but steel has two). The same ordering applies for words that have more consonants at their end (so, design and development rather than development and design, where design is pronounced with a single final consonant and development with two; or for and against, where against has 2 final consonants but for has either one or none, depending on the variety of English – American speakers, for example, usually pronounce the final /r/).

Mollin found that only 18 per cent of the pairs never occurred in the reverse order. These ‘frozen’ pairs were mainly those whose ordering was in terms of power. The vast majority of pairs, then, could be reversed, though most had a preferred order. The likelihood of a pair occurring in the reverse order depended on the type of factor governing the order, with pairs satisfying the semantic constraints of power and iconicity less likely to occur the other way round and the least reversible pairs satisfying both semantic and metrical constraints. This accounts for rich and famous, which seems to be almost completely frozen: famous and rich occurred only once in the BNC.  Knowledge and skills, by contrast, occurred just as often as skills and knowledge.

Mollin points out that once a certain threshold of frequency has been reached for the pronunciation of a pair in a particular order, ‘freezing’ becomes an inevitable process. This would mean that enough speakers had internalised the sequence in a fixed order, and therefore only produce it in that order, for enough other speakers to hear the sequence only in that order. They too would then pronounce the pair exclusively in that order and internalise it as a frozen sequence. In turn more and more speakers will hear the pair in only one order, until eventually it is a completely frozen sequence that never changes.


Mollin, Sandra (2011). Revisiting binomial order in English: ordering constraints and reversibility. Journal of English Language and Linguistics 16: 81-103.

doi. 10.1017/S1360674311000293

This summary was written by Jenny Cheshire

No comments:

Post a Comment

Note: only a member of this blog may post a comment.