Readers Write In #854: Naan thaan antha paiyan OR What’s in a name? OR Why I didn’t go out for two weeks

By Samyogita Hardikar

Ponniyin Selvan, especially the first one, was life-changing for me. It had everything that had been largely missing from my film-viewing life for over a decade. A soundtrack that sounds like nothing else you’ve ever heard, a setting that takes you to another world and makes you want to stay there, a plot and screenplay that (for the most part) respects, nay, demands your intelligence and attention, characters whose inner lives seem as rich as their costumes and jewellery, actors who seem to be born to play the roles they’ve been given, and a 100% Indian story with 0% of the cheap politics and brain-dead jingoism that you usually get these days.  I say life-changing, because looking back, I think it made me fall in love. With everyone who was involved in that film, and with Tamil (ahem… Tamizh) itself.  Who are all these wonderful people, I thought to myself. Why hadn’t I watched every single one of their films yet? And then I proceeded to do just that. One actor at a time, whatever I could get my hands on legally or illegally, then directors, and now just most new and interesting releases, like I used to do with Hindi films.

So, barring the odd dubbed Mani Ratnam or Kamal Haasan that made it to DD, my Tamil-watching age is around three years, or around 150 films. I started with Karthi and Vikram, and started noticing that a lot of their films were named after their own character. Same with all the Loki films, and a lot of the big releases that I heard of P.P.S. (Post Ponniyin Selvan). Just a one-word title, usually eponymous with, or descriptive of, the hero. So then I started wondering if this pattern was just a sampling bias on my part, a temporary bubble, or an actual difference between Hindi and Tamil films cultures.

The curious scientist part of my brain, which overlaps almost entirely with the really, really stupid time-waster part of my brain, wanted to look at some proper data. I thought I could automatize most steps of the data collection. I asked GPT to give me the collected filmographies of the top 20 film actors of the past 50 years, from Hindi and Tamil. GPT, gemini, and copilot all said that I was being unreasonable, and the best they could do is give me 20 names. I said FINE, but I want 20 from Hindi and Tamil EACH. GPT said FINE, and we both got to work. What I thought would be a simple project that I could knock off over a day, ended up taking two weeks in the end (Did I mention the really stupid time-waster part?). So, buckle up! Because I’ve suffered for my science, and now it’s your turn.

These are the top 20 that chatGPT suggested:

Hindi: Rajesh Khanna, Amitabh Bachchan  Dharmendra , Jeetendra, Shashi Kapoor, Rishi Kapoor , Vinod Khanna, Anil Kapoor, Jackie Shroff , Sunny Deol, Govinda, Shah Rukh Khan, Salman Khan , Aamir Khan, Akshay Kumar, Ajay Devgn, Hrithik Roshan, Saif Ali Khan, Ranbir Kapoor, Ranveer Singh

Tamil: M.G. Ramachandran (MGR), Sivaji Ganesan, Gemini Ganesan, Jaishankar, Rajinikanth, Kamal Haasan, Vijayakanth, Prabhu, Karthik, Sarathkumar, Arjun Sarja, Ajith Kumar, Joseph Vijay, Suriya, Vikram, Dhanush, Sivakarthikeyan, Jayam Ravi, Karthi, Silambarasan (Simbu/STR)

The only substitutions I made to the Tamil list after consulting two Tamil people were replacing SK (too new as a “top” actor) with Sathyaraj. And Jaishankar had to be excluded because his filmography on Wikipedia is just too incomplete when it comes to character names.

In Hindi, I substituted Rajesh Khanna for Dilip Kumar. Because Rajesh Khanna has around 300 films, and Hindi already had a bigger sample than Tamil, and because Dilip Kumar was a closer match to MGR and Sivaji Ganesan in terms of chronology and artistry.

I used the filmography tables available on Wikipedia. And removed all the non-Tamil films for the Tamil actors, and non-Hindi films for the Hindi actors. I also removed a lot of the films marked as “cameo”, but then started noticing that what people have labelled “cameo”/ “guest” or “special appearance” is very inconsistent. So gave up half way. I did try to remove all the roles marked “Himself” though.

This got me to 4,326 films from 39 actors. 2245 Hindi, and 2081 Tamil.

Then I divided the film titles into four categories:

Eponym: Film name = hero’s character. e.g., Vikram, Asoka

Descriptive: Film name = not a strict eponym but descriptive of or referring to the hero’s character. e.g., Coolie, Himmatwala

Group/ pair Eponym: Films name= more than one central character. e.g., Jodhaa-Akbar, Amar Akbar Anthony

Non-eponym: everything else.

When I split up the data between decades, this is the overall distribution of these four categories.

As I mentioned earlier, Hindi has a few more data-points than Tamil, and the distribution is also not even across decades, so in the next plot, let’s look at the same data as a percentage of different title types out of all the films from that decade. For both these plots, I removed all duplicates so that each film would only be counted once. E.g., Ponniyin Selvan would count as an Eponymous entry for Ravi Mohan, but would also show up as a Non-eponym for Vikram, Karthi, Sarathkumar and Prabhu. So for these plots, only the Ravi Mohan entry is considered, because the title IS eponymous, and should count towards that group if we want to see how many of all the films in 2020s were eponymously titled.

Except for the 40s and the 70s, Tamil films of “top” actors, compared to Hindi “top” actor films, have always had a greater proportion of Eponymous titles. The general pattern holds for most decades regardless of whether we consider only the “strict” eponyms, or also the loose-eponym categories (descriptive and group/ par eponyms). 

But we still don’t know whether it’s only a few actors driving this difference, or a more generalized tendency in Tamil. For that we can plot data from each actor separately. For this plot, I went back to the datasheet with “duplicates” in it. To go back to the example of Ponniyin Selvan, in the plot below, it’s included not just as an Eponym for Ravi Mohan, but also as a non-eponym for Vikram, Karthi, Sarathkumar and Prabhu.

Note: The Rishi Kapoor datapoint of 100% in the 2020s, is sad and sort of unfair, because that was just one posthumous release in 2022, and his parts in it had to be completed by Paresh Raawal.

We see a pattern very similar to the bar graph. The Tamil panel has many more points in the upper half of the graph (more than 50% eponymous titles in a given decade). And the proportion is much higher in both groups from 80s onwards. Rajinikanth, Vijay, and Karthi stand out immediately as having spend a lot of time in the upper half of the graph.

Next, I wanted to see how this distribution changes depending on how many films you’ve done. This is partially a proxy for age/ career transition (more films= longer career= probably greater age and greater number of character roles), but also, leaving the working-speed of the 70s and 80s aside, it’s much easier to do ten non-central roles in a year than ten hero roles. So, you would expect to see a negative correlation between number of films.

And that’s what we see. You can draw a negative correlation slope over these data points. But there are two sub-patterns here. The cluster at the bottom marked by Hrithik, Vijayakanth, Sivaji, and Gemini at the vertices. This is more or less a normal distribution, people who have done anywhere between 30 and 300 films, with less than 25% eponymous titles. And another cluster in the top left part of the graph- actors who have fewer than 75 credits to their name and 20-50% eponymous titles. It makes sense that fewer credits= more eponyms. But what’s interesting is that these are all post 90s actors. 

It could still be the case that this cluster is just showing the early peaks in a top actor’s career when you do you most “hero”-ic films. (Also, Sorry, Vikram! Had to omit the 20 or so Malayalam/ Telugu films from your struggling days, so the stats are unfairly inflated compared to the rest). So what we want to see is the cumulative percentage of eponyms over time.

What we see is that four actors from the top-left cluster Ranveer, Ranbir, Ravi Mohan and Karthi, (and to a lesser extent Dhanush and Vijay), also stand out when we consider entire career trajectories. Unlike most other, they didn’t rise to their peak over the first 30 films. They started at the peak, and have stayed in the higher percentages for the most part.

The uniqueness of their trajectories becomes even clearer in the graph below when you restrict the data to just the first few films from each actor.

But what’s even more interesting is that Ranveer and Ranbir are in that part of the graph too. If you go back to the very first barplots, it’s obvious that in the 2020s the difference between Tamil and Hindi has almost disappeared. The testosterone level in big Bollywood movies, especially compared to the super-sweet, duper-romantic Khan years that I grew up in, is proof enough that times have a-changed. But let’s look at one last plot to see what, or who has changed the most. This time I picked just the top, top actors who are still doing “hero” roles. It’s tricky to do one on one matching between Hindi and Tamil here, because Rajinikanth and KH are still doing mainly central roles, but Amitabh Bachchan is the only one from Hindi that would make for a fair comparison with the two of them, in terms of epoch, career length, talent and stardom. The rest I picked based on what kind of scale their films tend to be made at, but that’s a fickle metric, so let’s take it with a pinch of salt.

Amitabh, probably the biggest star Bollywood has ever seen, definitely seems to have settled into non-self-focused roles post 2000s. Rajini and Kamal, who are closest to him in age, have not. The Khans (especially Salman), Hrithik, Vikram, Vijay, Dhanush, Ranbir have all ramped up the me-role machine, and with the exception of Ajith (and to some extent Suriya), so has everyone in the Tamil panel. But Hindi is definitely doubling down way more than Tamil.

Now, I know that this is a very narrow look at what’s going on and there are lots of caveats in terms of what the data is, how it’s been coded, and how much meaning we can really assign to these patterns. After all, a title is not a perfect proxy for the content of the film. The film that started all of this for me, Ponniyin Selvan, a seemingly perfect “Eponym”, is hardly at all about Ponniyin Selvan or Ravi Mohan. And “Dangal”, a perfect “Non-eponym” ended up doing the great Indian Bollywood trick in the last 15 minutes so that Aamir Khan’s character would still get to be the central one. But I do still think that the pattern is telling, and I hate that we’re heading in this direction. Here’s hoping that we change course very soon.

I also simply chose to take the top actors because “follow the money”. The patterns would probably look slightly different if we took all the films of the last few decades. But GPT (AND the documentation of our films in general) needs to improve a hell of a lot more before I can take on anything like that.