The plausible origin of ‘thunder and lightning’ has been dealt with in Graffitto's answer, but I want to counterbalance the impression that the binomial thunder and lightning is Biblical. The expression may have made its first appearance in the Exodus, but that is not sufficient evidence to explain why the English use that fixed combination.
It is generally accepted that the first English Bible printed in England was written by WilliamTyndale in 1526. It was also the first time the Bible became available to the secular population, thus for the purpose of this question, any translation of the Holy Bible prior to that date should not be taken into consideration. After Tyndale’s death in 1536, the Bible was reviewed several times until its most acclaimed version, the King James Bible, was finalized and printed in 1612. The task of translation was performed by over 40 Church of England scholars; the New Testament was translated from Greek while the Old Testament was translated from Hebrew and Aramaic. The Book of Exodus is found in the Old Testament.
The expression thunder and lightning is not cited verbatim in the King James Bible. The precise citations in Exodus 19:16 and 20:18 are the following
And it came to pass on the third day in the morning, that there were thunders and lightnings, and a thick cloud upon the mount, and the voice of the trumpet exceeding loud; so that all the people that was in the camp trembled.
And all the people saw the thunderings, and the lightnings, and the noise of the trumpet, and the mountain smoking: and when the people saw it, they removed, and stood afar off. And they said unto Moses, Speak thou with us, and we will hear: but let not God speak with us, lest we die.
Which means the people heard the thundering and then saw the lightning. The one followed the other. The expressions, thunders and lightnings and thunderings, and lightnings, were later reduced to the more familiar thunder and lightning which we hear and use today. But why do English speakers prefer this fixed order? When today we know that lightning precedes thunder.
Sequence and Binomials
Joe Blow's answer included the binomial Donner and Blitzen, (now deleted) and argued the English thunder and lightning scanned better. However, a rebuttal might claim the Germans copied the order from the Holy Bible.
Is the same therefore true for all European languages? Not really. If we look at Google Ngram we see that the German, Donner und Blitz, and Blitz und Donner are both common, but the latter (lightning and thunder) appears to be increasingly dominant, I might rationalize that this is due to semantic constraints, i.e. lightning precedes thunder, but I digress. In Spanish, trueno y relámpago and rayos y truenos are both used; and in Italian the swapped binomial: lampi e tuoni (lightning and thunder) is often heard too. In fact, Google Italian Books lists 4,240 instances. Janus Bahs Jacquet, in the comments, mentions that the Swedish, blixt och dunder is a fixed expression, and the Danish, lyn og torden is completely fixed in the opposite order of the English phrase. Which begs the question, why didn't all these European languages follow suit? A possible explanation is to be found in the following sources.
Binomials, the pairing of words, has been mentioned several times on EL&U:
- "pros and cons", "black and white", "ups and downs". Always in a fixed sequence, is there a word or phrase for these?
Binomials have two main characteristics. The first, […] is that the order is usually perceived as fixed. The second is that the two terms are normally the same part of speech, though not always Nouns. […] Binomials occur in many languages and are often the subject of academic research.
- Why do we say kith & kin and not kin & kith?
Cooper and Ross give the following hierarchy for words that differ only in their final consonant (in those pairings where there is no ordering based on semantics for various reasons). Words that end with the types of consonants further to the right in the list below are likely to be in the final position.
- Stops - Aspirants - Nasals - Liquids - Glides
Basically the more obstruent-like the final consonant the more likely that word is to be first in the pairing.
- crisscross, dillydally, riffraff, etc
Cooper and Ross* note that there are a number of reasons, some phonological and some semantic, that explain why one does not find reverse freezes like *raffriff, *pocus-hocus, *crosscriss, *dallydilly
Which word comes first in an irreversible (binomial) can be dictated by what is considered the most powerful or important by its native speakers: e.g. Adam and Eve; bride and groom; law and order; food and drink.
The number of syllables. Often the first word will be singular and have fewer syllables, e.g; bow and arrow; horse and carriage; rich and famous; cease and desist; milk and honey; smoke and mirrors; ladies and gentlemen.
Alliteration, e.g; flip-flop; right or wrong; spic and span; sugar and spice.
Semantic contraints, words which follow a natural logical order, e.g; sooner or later; now or never; start to finish; life and death.
Phonology. Among the too many to mention, the word that contains the voiced consonant will precede the word containing the voiceless consonant.
The term lightning is represented in American IPA as [ˈlʌɪʔ.nɪŋ]. The stress is on the first syllable as shown by the apostrophe. The symbol ʔ states that the letter t is a glottal stop, (the sound in the middle of the word uh-oh) when it comes after a vowel and it is followed by an n sound. Listen to the British pronunciation of lightning on Macmillan Dictionary, which is virtually identical to the American English pronunciation, for confirmation.
Whereas the letter d in thunder is plosive, and voiced. Linguists call it a voiced alveolar stop. The IPA for thunder is /ˈθʌn.dɚ/, note the same stress on the first syllable as in [ˈlʌɪʔ.nɪŋ] and that both words have two syllables, which invalidates Joe Blow's assertion that lightning is the longer word (the "shorter" scan word going first), it is nothing of the sort.
It is my supposition that thunder comes first in the binomial because of phonological constraints, i.e. the voiced d whereas in and, the d is barely perceptible and the word is often pronounced as n. The end result is something like: thunder'n'lightnin', which is very similar to the irreversible binomial and alliterative rock'n'roll, where the letter k has a ‘harder’ sound than the letter ‘l’.
Source: p74 (4.1.4) p85-87; The (Ir)reversibility of English Binomials: Corpus, constraints, developments By Sandra Mollin