Okay, you've decided on some sounds that you want to use in your language, on a way to write the sounds, and on what sounds can go where; now it's time to start word-building.
Word building isn't as simple as picking a word from the dictionary and making up a pattern of sounds (don't worry, it's not as boring either). If you look in an English dictionary you will find many words that can be broken up into smaller words. For example, the word everybody can be split into the two words every and body. Even more often, you will find words that can be broken up into pieces that aren't actually full words. Indescribable can be broken down into in-, describe, and -able. You will notice that in- and -able, while not full words, still mean something (even if what they mean isn't exactly clear); the same word-parts appear in inadequate and readable and many other words.
These 'half words' are called affixes, and they build new words from simpler words or based on stems which do not stand alone (like in Latin, stems are a part of words, like "cor-"). (There are also affixes that change properties of a word without changing what word it is, like English -s for plural. The inflection, like plural "s", can be applied using affixes, but we'll deal with those in the next section.) You'll almost certainly want some word-building affixes for your conlang. They give your vocabulary texture, and the more of them you have, the more words you can add to your vocabulary just by inventing one new simple word.
There are lots of different kinds of affixes, depending on how they change a word when they attach to it. Some simple techniques you can use:
- add onto the front of a word (like in-).
- add onto the end of a word (like -able).
- insert into the middle of a word (like hip-hop -iz- in house → hizouse).
- change part of a word (like goose → geese; of course that affix doesn't build a new word, but it shows another way an affix can work).
An affix may take different forms on different words, like -able that on some words becomes -ible. Sometimes the forms used on different words are so different that you might think of them as different affixes, that just happen to mean the same thing and apply to entirely different sets of words — like the plural affix in goose → geese, where for most English words the plural affix has the form -s or -es.
If you set up your word-building affixes to work the same way as in English, even if they use different sounds and attach differently, they'll still make your conlang feel similar to English. For example, if you have an affix equivalent to English un-, don't use it exactly the way English does. Experiment with it — you might have no word for "bad", but have "ungood" instead; maybe have a word for "uncommon" that doesn't use the "un-", in the same way "bad" isn't "ungood".
The English word manual comes from the Latin word manus, which means "hand". Right? Well, sort of. Latin is a language with relatively few different words, so each word has to do a lot of work. The Latin word manus does mean "hand", the part of the human anatomy; but it also means "fist"; "handwriting"; "finishing touch"; "workmanship"; "team"; "gang"; "band of soldiers"; and "trunk", the part of the anatomy of an elephant. The English language has a tremendous number of different words, so the average word doesn't have to do nearly as much — but especially basic and common words may still have a fairly wide range of meanings, and the English word hand is like that; it can mean everything from a person's handwriting (like in Latin), or the pointy bits on a clock face, to the set of cards that a player holds in a card game. (Compare: Latin manus, English hand.)
No word in one natlang means exactly the same thing as any word in another natlang. To keep your conlang from feeling like English, don't give it words for just the same concepts as in English. Try to invent words that have the meanings of more than one English word, or are more specific than English words. Even better are words that fall between levels of description in English. For example, you might create a word that means "a four-legged hairy animal", to cover dog as well as horse, but not as much as mammal or animal. Think about a whole set of related or similar things (for instance, colors, animals, or family relationships), and how your language can divide that mass of reality into manageable parts in a different way than English. And think about additional, perhaps figurative, senses of a word (like all those extra meanings for manus or hand).
You can even create words for concepts that are only very clumsily expressed in English, such as: that awkward situation when something has happened, which is known to more than one present party, which neither wish to speak about, and suddenly they have nothing to say to each other. Another word could mean the same thing, but with another, unknowing party present, who senses a weird atmosphere between the other parties.
Part of the challenge of conlanging — whether you're building an artlang, engelang, or auxlang — is that you won't really know how your language is going to feel until you've got enough of it together to start saying interesting things in it. (That's also one reason why conlangers may tend to keep tinkering with their creations.) So how can you tell, early in building your vocabulary, whether it will come out too much like English? Well, as a rule of thumb, building a vocabulary of a hundred words or so should be a pretty massive job — done properly, it's a major milestone, like a small company landing its first big client — and each additional hundred words should be heavy work as well. If you easily reach, say, a thousand words, you're probably in real danger of producing a conlang that will always feel too much like your native language, no matter what sounds and grammar you give it.