Naming and NPCsity

Last post discussed player-directed – or, at least, not directly author-mandated – naming for individual player characters. But there’s another important use-case: naming lots of characters, using generated names.

The Simple End of Things

Creepy narrative block-puzzler Road Not Taken has a town populated by a half-dozen NPCs whom you can befriend or romance. They’re somewhat randomised each time – their colouration changes, and so does their name (they only have first names). Each of them retains their individually-crafted conversation and preferences, so they feel basically like versions of the same person. And while their names are randomised, they each have their own list of possible names, so they always retain a similar feel.


Alas, poor whatstheirface.

This is a modest amount of randomisation, but it’s about right for the piece; because a lot of the interest of the characters comes from accumulating evocative hints about their personal tragedies, it’s unlikely that they would work without pre-scripted sets of responses. The randomisation suggests a certain kind of recurrence, of the same roles filled by new actors.

This is an easy-to-manage system, but it has odd effects and works mostly because Road Not Taken is an odd game that wants you to feel uneasy about these relationships. Hollywood Visionary might be a less specialised example: it allows the player to choose or randomise the names of many of its major NPCs. The most obvious effect of this is to aid distinction between individual playthroughs; while the name-change neither reflects nor causes mechanical changes to the character, it acts as a distinguishing feature to keep different sessions of the game distinct.


More specifically, it strengthens the sense of each iteration of that NPC as a distinct character, which is particularly important when the plot rests on relationships with them. When I replayed Choice of the Deathless and discovered that romancing very different NPCs led to the same (highly abstracted) sex scene, my response was approximately wow, you didn’t even change the sheets. To unpack that a bit: things that emphasize the interchangeability of characters tend to weaken our sense of their integrity as characters, and the replayability of games is a big hazard to that. If we get the impression that Assistant Dottie will dutifully fall into the arms of any PC who bothers to pay attention to her, Dottie’s arc begins to seem less like a personal relationship and more like a vending machine. Changing the name may not change the type of Dottie, but it emphasises her as a distinct token, and that is kind of a big deal in how we think about individuals.

Much of the time when games assign names at runtime, however, it’s because the characters themselves are the product of generation. That’s a tougher nut to crack; let’s begin with a gorilla.

Generated Names For Generated People

As is its wont, Dwarf Fortress is spectacularly detailed about its naming system. Dwarves have a constructed language – a vocabulary, at least, if not a complete grammar – and construct names out of compound words. The game sometimes displays the English translation, sometimes the Dwarven. Dwarven vocab has a consistent phonology, so the names all feel like the same language – but even the English translations have their own flavour. Part of this is because of the content of their vocabulary: just as we can be fairly sure that the speakers of Proto-Indo-European were familiar with wheeled vehicles and honey-bees, you can tell that dwarf culture has a focus on construction, industry, physical geography, warfare, certain kinds of decorative arts. When you want to name your fortress, you have to draw it from the Dwarven vocabulary and construct it according to Dwarven naming grammar: a cumbersome process, but one that allows both tonal consistency and creative freedom for the player (even if it’s likely to be used for a dick joke). Broad, yet circumscribed: a pity that it’s not more usable.

DF uses much the same system for the names of people, though the player doesn’t get to choose those. For ease of recognition, this isn’t so great. Dwarf names are all very similar and have few distinctive connotations to an English ear, so they work poorly as a means of quick identification. In some contexts, dwarf surnames are presented in Dwarven; in others, in the English translation, which adds yet more information for the player to get overwhelmed by.

As Gabriel García Márquez and Dr. Seuss illustrate, names become difficult to use when they’re insufficiently distinct. In the real world, surnames tended to arise as bureaucracies arose and needed more distinctive and stable identifiers for tracking people and their relationships: but game populations rarely exceed that of a village, where more casual and distinctive names are more efficient. Dwarves can be granted player-entry nicknames, and I almost always need to use this, at least for the more important members of the fortress. I can never get to the point of remembering dwarves by name. This has a great deal to do with the extent to which dwarves function as individuals: while every dwarf has detailed and unique characteristics, these are all generated content. And you can have hundreds of dwarves in a single fortress.

To recap: when you’re presented with a big pack of NPCs, the goal of consistency (names as a whole contribute to the sense of a consistent world) often clash with the goal of distinctiveness (names help you pick out and remember individuals). The Dwarf Fortress approach is heavily weighted towards consistency.


Does it count as a diversity hire if his main role is to be Guy Whose Name I Won’t Confuse With Anybody Else’s?

The NPC party members in Expeditions: Conquistador have Spanish names (at least, the initial ones do). Naming in the era was fairly conservative, so they all follow a pretty similar style; and I lack much of the cultural background that would bring specific connotations to each name. Early on, the names weren’t much help as identifiers, particularly since the game’s models for characters in battle are not strongly distinctive either. But these aren’t generated characters; and the game treats your NPCs as individuals, with unique portraits and scripted character events. After a while my ear attuned and the names took on more meaning; but it took a bit longer than it would have under the classic Fellowship-type party, where each character comes from a different culture with its own naming style.

Battle for Wesnoth, a Tolkienian-fantasy turn-based strategy game, generates names for its non-unique, recruited units using Markov chains seeded with lists of example names for each race. This has some advantages – it lets modders create new naming styles without having to construct and code a bunch of phonological rules, because Markov chains infer rules for you: all you have to do is input a set of example names. Markov chains often produce weird results, but it’s good enough for fantasy-language mooks. The system has limited usefulness, though, because the names end up too stylistically similar: they’re not very helpful for quickly distinguishing between a bunch of otherwise-similar elves. I generally end up renaming my troops with role-related joke names, Firebert for a red mage and Twinklebritches for an elf shaman; Wesnoth doesn’t really have much cultural depth, so my motivation for thematic consistency is not strong.

Themed Party Renaming

OK, let’s talk about that renaming thing for a minute, since it’s come up a few times. It’s pretty common, in games with multiple NPCs under partial player control, for the player to be given list-generated names with the option to change them. The most famous example of this is in Oregon Trail, in which everybody at some point names their game family after their own family and friends. And XCOM-playing friends delight in telling me “oh, you’re a sniper, but you haven’t leveled up much because I mostly use my grandma.”

There’s obviously something gained here from a funny, unserious linkage to the real world, importing familiar meaning even if it doesn’t make much sense. There’s more going on here than just silliness, though.

The replay issue looms large for XCOM; it always has essentially the same story with the same plot beats, and it recycles maps a lot. Distinguishing one playthrough from another takes some mental effort, and failing to do so can have consequences both in-game (you assume you’ve already taken care of it, but you haven’t) and in experiential, narrative aspects (one game-generated narrative gets muddled up with another; the game elements seem less individual and more expressions of a general type.) Emergent narrative – where a player imaginatively constructs a story from play that is more than the sum of its parts – relies on the distinctive, and is strongly facilitated by memorable naming.


BOOOORING. Your name is now May Murrumbidgee.

 Naming systems can serve as mnemonic devices, picking out a specific playthrough against its peers, distinguishing one randomly-generated mook from another. Often this means picking a new naming-system for each game. For a while in XCOM I’d name my soldiers after features of their home nations – first name from a literary figure, surname from a river, that kind of thing. The exact method didn’t matter so much – what was important was that the names were memorable enough for game purposes. Almost always, such names are less plausible than the John-Smith names the game generates; utility trumps realism here. But I’m not doing it purely for the gameplay advantages of recognition, for remembering which Assaulter has the stun-gun: in helping me to hold characters in my head as individuals, it lets me conceive of them more like individual characters in a particular story. The mnemonic and narrative functions are closely-aligned here, and aligned against world consistency.

XCOM has been discussed as a superhero game in denial – and this is definitely related territory, because superhero fiction has long used punchy, memorable naming of mundane alter-egos (together with iconic costumes) as a mnemonic device to distinguish characters whose faces, bodies and personalities a) are often very similar and b) change around as writers and artists come and go. I think there’s a case to be made that if you’re writing a name generator, you shouldn’t usually be aiming for typical names so much as superhero-type names – alliteration and assonance, strong poetic feet, avoiding territory already namespaced, possibly even some reflection of role. This is a way more complicated exercise than shuffling two lists and jamming the results together.

In Dwarf Fortress, where there are far more controlled NPCs and I need to pay less regular attention to them individually, I tend to give my dwarves nicknames based on their profession: Hopyard, Codguts, Stabby. That’s the information I need most often in a hurry. This tends to reduce dwarves – who in mechanical terms are much more complicated, individually distinctive beings than XCOM soldiers – to their mechanical roles, minimising the distracting, unimportant (and overwhelmingly numerous) details of individuality. If I was turning a DF game into a written narrative – at least, beyond the scope of ‘isolated funny incident’ – I’d feel uneasy about using those handles, because so much of the narrative of DF is about language and history.

The Crusader Kings games offer a middle ground, albeit one that only works because of its specific interest in modeling culture, family relationships and inheritance. Normally, a child is given a name randomly selected from a list determined by culture and gender; but there are refinements. There’s a decent chance that a child will inherit its name from a parent or grandparent, which reinforces the sense of dynastic continuity. Some names are tied to specific religions, so that (e.g.) Christian Arabs won’t name their children Muhammad. In CK1, children sometimes got names based on the culture of the province their parent was in at the time of their birth. Names have cross-cultural equivalents – so if a Norse Eiríkr has an English grandchild of the same name, he’d be Eric. (This is a super-nifty feature that, alas, gets used rarely and noticed less.) Certain deeds can earn rulers titles on the end of this.

CK2’s naming is well above the average on consistency within its world: the flavour of the names is a big deal for how you experience each culture. You can rename your own children when they’re born, but I rarely do; it doesn’t seem right. But doesn’t always do as well at recognition and memory. Some of this is because it can offload its recognition difficulty onto portraits; the name doesn’t need to take all the weight. It suffers some of the same problems as Dwarf Fortress, too: there are hundreds of named characters and you honestly don’t need to pay attention to all of them. One of the major differences between CK1 and CK2 is that the latter plays smoother, easier, faster; there’s less need for careful attention to detail, and this makes it more likely that you’ll play without spending as much time picking over the individual details of characters, getting to know them, developing hopes and ambitions about them.

Anyway. Next time I’m going to talk at length about a giant, tower-of-Babel name-generator that I had a lot of fun failing at.

This entry was posted in cyoa, interactive fiction, videogames and tagged , , , , , , , , . Bookmark the permalink.

1 Response to Naming and NPCsity

  1. Pingback: Hollywood Visionary | These Heterogenous Tasks

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s