All the time in the world…

Measuring the effect time has on delivering a great speech using data from Toastmasters

Toastmasters
Time
Public Speaking
Analysis
Author
Affiliation

Vadym Yudenko

Published

August 4, 2023

Modified

November 6, 2023

Introduction

Consider time as the invisible thread weaving through the fabric of every conversation, the silent drumbeat to which all effective communication must sync. It’s not just about being on time — though punctuality certainly speaks volumes — it’s also about respecting the rhythm of discourse, whether it’s a brief exchange or a lengthy dialogue.

Imagine the different scenarios: a toast at a family gathering, where brevity is laced with warmth; a debate, where each second is a battle for conviction; an elevator pitch, where within a minute, careers could be made or opportunities lost. While these situations differ wildly in content and context, one rule remains universally unchallenged: overrun your allotted time, and your message may fall on deaf ears — or worse, you might find yourself ‘toasted’!

Moving beyond anecdotal evidence, we dive into the realm of data. To understand the practical implications of time (being in time, on time, overtime ) in communication, we’ve turned to a domain where time isn’t just a factor — it’s a contestant: the Toastmasters International meetings, specifically Table Topics session!

If you’re familiar with Toastmasters International, then jump straight to the results. If not - keep reading.

Toastmasters in a Nutshell

Toastmasters is a place to practice your public speaking skills. Each meeting, which typically lasts for 90 minutes, has a structure:

  1. Prepared speeches session: usually 5-7 minutes long speeches on various topics, these speeches are for members of the organization;

  2. Table Topics (TT) session: 2 minutes log impromptu speeches, everyone can participate, and the best speech is voted for;

  3. Evaluation session: everyone who came on stage during previous sessions is given feedback on how well they did.

Most Toastmasters members are working professionals from different fields who want to network, be confident presenting, make new connections, and have a great time. Delivering prepared speeches allows members to structure their thoughts and practice their delivery. The Table Topics session, on the other hand, challenges members to speak impromptu. In this session, you can either volunteer to answer a question or be selected by the Table Topics Master. The essence of this activity is to encourage speaking “on your feet” rather than from the comfort of your seat.

During the Evaluation session a Timer of the meeting - a person specifically assigned to the timetracking task - presents their report, sine qua non1 of this article. Timer’s responsibility is to keep track of time and indicate to the speakers how much time they have left before they have to wrap up. Without going into much detail, green is an early warning, yellow - soon the time will be up, red - time is up, finish your speech. At the end of the meeting a time has a report of speakers and their timings

  • 1 A necessary condition without which something is not possible [from Latin]

  • Note on Timer reports

    Timer tracks time of each speaker, as can be seen on the reports. Reports usually have the following information useful for our analysis:

    1. Speaker Name
    2. Speech Time
    3. Club
    4. Date
    5. TT winner


    Data

    Like a famished mouse looking for breadcrumbs under the baker’s table, the author was open-sourcing these reports from Telegram , Facebook , some were written by author himself and other were transcribed from videos of the meetings.

    Most of the data was collected from Ukrainian TMCs: Chamber2, Arttalkers, Changemakers, Dniprohills, KOMA (UA-speaking) and Svoya Sorochka (UA-speaking). There are some reports from Japanese and Polish TMCs, but we will not be focusing on them.

  • 2 Special thanks goes to Chamber and their Timer reports, without them this project would’ve taken much-much longer.


  • In total there are 180 unique meetings in the dataset from 2018 till 2022. Some of the meetings were dropped due to NAs in at least one of the columns. Speaking of gender discrimination, there was none present, as the number of men and women3 is equal.

  • The gender of the speaker was assigned post hoc, based on the reports/photos of the meeting, common sense and a little bit of prejudice :)

    In cases where it was impossible to establish the gender of a speaker - the whole meeting was excluded.

  • Gender # of TT paticipants Proportion, %
    Female 682 47.63
    Male 700 48.88
    NA 50 3.49

    With regards to TT participant’s names there are a total of 312 unique names 4.

  • In reports the same people were reported under slightly different names. For example: Alexey/Aleksey/Alex/Oleksiy/Oleksey…. and their counterpart Cyrillic versions. To deal with this issue, all versions were renamed to Alex. Similar approach was taken with other names.

    All Ukrainian names were translated to English.


  • And lastly we have Time of Table Topics (TT)! How TT works:

    You have 2 minutes to answer a question or speak on the given topic. To stay within time limits, the timer will show you green sign when there is 1 minute left, yellow when 30 seconds and red when time is up.

    Ideally you want to have an opening-body-conclusion short speech, but this isn’t always possible. You can volunteer or be dragged on stage. The question may be familiar to you or something you haven’t even thought about. That is all to say, there are many different things happening during TT sessions and winning it requires achieving many things. At the end of TT everyone votes for “the best speech” - this is our dependent variable.

    Question!

    Before we get to figuring out what it takes to win TT, given the information above how does the distribution of TT times look like? A, B, C or D?

    Surprising? Either way, it is peculiar how these times fall into one of nature’s patterns!

    Analysis

    In this part we get to know more about our dataset, whilst answering some of the hardest philosophical questions! Starting with…

    Who likes to talk more, women or men?

    
    Regression results
    ============================================================
                            Individual Time         TT Time     
                                  (1)                 (2)       
    ------------------------------------------------------------
    Gender                       2.372              -0.347      
                                (2.071)            (28.473)     
    Ratio of male/female       -15.199**          -585.821***   
                                (6.131)            (84.311)     
    Observations                 1,366               1,366      
    R2                           0.005               0.038      
    Adjusted R2                  0.003               0.037      
    ------------------------------------------------------------
    Notes:               ***Significant at the 1 percent level. 
                         **Significant at the 5 percent level.  
                         *Significant at the 10 percent level.  

    The answer? The difference is not significant! The distributions closely resemble each other. Women don’t speak longer than men!

    Women actually speak a little less than men. However, we cannot say that women ACTUALLY talk less! In the current study we only observe time taken, but within these 2 minutes on stage, they might say more (on average) than men. There seem to be reports that women communicate ideas differently and use more words like ah/uh/m-m-m/you know/so/like5.

  • 5 Foreshadowing a future ah-counter perspective on winning TT?!

  • Strangely enough, if the proportion of men in TT session increases - then TT session will be be shorter. Also, having more men speak during TT session is associated with lower TT times.

    Do we speak longer online?

    In-person conversations and presentations require you to engage with people directly; you see, hear, feel, and even smell them throughout your speech. There are countless non-verbal signals that are communicated without us even noticing. During face-to-face meetings, if you take too long to finish, people may start fidgeting, yawning, or looking away. This indicates discomfort, prompting you to wrap up your speech quickly. In contrast, when interacting online, you are dealing with avatars, and the only means of gauging your audience’s reactions is by observing their responses. This may lead to a tendency to go overtime more often than if you were delivering the same speech in person. We will explore this further below.

    
    Regression results
    ======================================================
                      Individual Time         TT Time     
                            (1)                 (2)       
    ------------------------------------------------------
    Online/Offline         0.580            -178.118***   
                          (1.916)            (26.812)     
    Observations           1,432               1,432      
    R2                     0.000               0.030      
    Adjusted R2           -0.001               0.029      
    ------------------------------------------------------
    Notes:         ***Significant at the 1 percent level. 
                   **Significant at the 5 percent level.  
                   *Significant at the 10 percent level.  

    Whether the meeting is online or offline does not affect the length of TT speeches. However, there is a significant association with the total duration of TT sessions: on average, online TT sessions are longer than in-person ones by approximately 160 seconds (equivalent to about 1.5 speakers). One potential explanation for this is the influence of the global pandemic. Now, participants do not need to be physically present in a specific location to join a meeting, and, perhaps most significantly, there is no commute involved afterward—one is likely already home.

    While the distribution of individual speeches remains “Normal” throughout the years, TT sessions became longer and take up to 40(!) minutes.

    
    Regression results
    ====================================================
                    Individual Time         TT Time     
                          (1)                 (2)       
    ----------------------------------------------------
    Year2012            34.977**            -130.545    
                        (16.572)           (217.757)    
    Year2013            32.583**            -100.583    
                        (16.279)           (213.902)    
    Year2016           45.750***             43.500     
                        (16.918)           (222.294)    
    Year2017            36.850*              -1.000     
                        (20.332)           (267.164)    
    Year2018           42.788***             73.423     
                        (14.420)           (189.472)    
    Year2019           39.441***            321.941*    
                        (13.331)           (175.163)    
    Year2020           43.136***           402.828**    
                        (12.794)           (168.107)    
    Year2021           44.867***           467.536***   
                        (12.719)           (167.132)    
    Year2022           46.424***           750.497***   
                        (12.699)           (166.861)    
    Constant           61.750***           494.000***   
                        (12.610)           (165.688)    
    Observations         1,432               1,432      
    R2                   0.013               0.155      
    Adjusted R2          0.006               0.150      
    ----------------------------------------------------
    Notes:       ***Significant at the 1 percent level. 
                 **Significant at the 5 percent level.  
                 *Significant at the 10 percent level.  

    Are there differences on the club level?

    Further exploring the data, we look at whether some of the differences can be explained by looking at distributions at the club level.

    Individual times don’t seem to vary that much across clubs, however KOMA (Ukrainian-speaking club) is an exception. As for distribution of gender of participants one can see the clubs that have more women or men participating.








    How to win Table Topics session?

    In this last section we will find out what it takes to win TT session!

    First of all, what doesn’t it mean “to win”? Winning means - majority voted for you and not someone else. Win can be attributed to 2 types of factors: internal and external. Internal factors can be broken down into articulation and story. Articulation is how well you communicate your idea/story: your vocal variety, vocal clarity, hand gestures, eye contact, stage movement, being timely, overall appearance, among other. Story has to do with your message to the audience. How clear is it, does it evoke emotion or make one think? Is there a flow to the speech, is it easy to follow?

    External factors can also be broken down into 2 categories: perception and environment. Perception is how your message is understood by the audience. This has to do with who those people are, what they find funny, what have they eaten before the meetings etc. Environment is the place where you give a speech. Zoom, in person, microphone quality, distance to the audience, acoustics…

    Flowchart made with ggflowchart

    What plays the biggest role in winning TT?






    Next we look at a set of hypothesis:

    Hypothesis 1:

    The last speaker has a higher probability of winning TT

    
    Regression results
    =========================================================
                                  Individual Time            
                           logistic             probit       
                              (1)                 (2)        
    ---------------------------------------------------------
    Speaker Order            0.009               0.005       
                            (0.033)             (0.018)      
    Observations              679                 679        
    Log Likelihood         -278.435            -278.435      
    Akaike Inf. Crit.       560.871             560.871      
    ---------------------------------------------------------
    Notes:            ***Significant at the 1 percent level. 
                      **Significant at the 5 percent level.  
                      *Significant at the 10 percent level.  

    Waiting and speaking later during TT session does indeed increases the probability of winning by a whopping 0.9%!

    This is a great news, because you can’t expect to win by just waiting and giving a speech at the end of TT session. There a lot that goes into giving a great speech and a lot more into winning, but speaker order doesn’t seem to be it. But one thing is clear here - the only people who don’t win are those that don’t try!

    Hypothesis 2:

    The likelihood of winning decreases as the length of the speech increases

    Since we have a dozen of TT winners who spoke over 150 seconds, we will remove them from our analysis and run regression with and w/o outliers.

    
    Regression results
    =========================================================
                                        TT                   
                      Model with outliers  Model w/o outliers
                              (1)                 (2)        
    ---------------------------------------------------------
    Individual Time         0.008***            0.013***     
                            (0.003)             (0.005)      
    Constant               -2.680***           -3.166***     
                            (0.343)             (0.497)      
    Observations              679                 609        
    Log Likelihood          -274.566            -243.462     
    Akaike Inf. Crit.       553.133             490.923      
    ---------------------------------------------------------
    Notes:            ***Significant at the 1 percent level. 
                      **Significant at the 5 percent level.  
                      *Significant at the 10 percent level.  

    Interestingly that probability of winning TT is increasing with every second! Even removing outliers (150 seconds or higher) doesn’t change the significance. You still better speak for as long as and not a second more! Otherwise you’ll be disqualified!

    Think about how this finding plays out in your life.

    When you have someone’s attention, you have to be mindful of time. You are not given their attention indefinitely, you have to use this opportunity exhaustively, making every second count. Speak fast and blurt out your thought and they will not understand it fully. Speak too slow and they might switch off and think that you can’t properly articulate your thoughts. Place yourself in between these 2 extremes and finish by the time they lose focus!

    Hypothesis 3:

    Time of other speakers has no effect on probability of winning

    
    Regression results
    ================================================================
                                               TT                   
                             Model with outliers  Model w/o outliers
                                     (1)                 (2)        
    ----------------------------------------------------------------
    Individual time                0.009**             0.013**      
                                   (0.004)             (0.005)      
    Time of previous speaker        0.000               0.001       
                                   (0.004)             (0.004)      
    Time of next speaker            -0.006             -0.007*      
                                   (0.004)             (0.004)      
    Constant                      -2.318***           -2.570***     
                                   (0.588)             (0.740)      
    Observations                     512                 460        
    Log Likelihood                 -196.804            -177.511     
    Akaike Inf. Crit.              401.608             363.022      
    ----------------------------------------------------------------
    Notes:                   ***Significant at the 1 percent level. 
                             **Significant at the 5 percent level.  
                             *Significant at the 10 percent level.  

    We can’t always give our best speeches, especially impromptu. Sometimes what beats a competition is competition itself. You can never underestimate your opponents, even when it fails miserably. Restricting our analysis to length of impromptu speech, the failure would be not using enough time (see H2). And our foes failing miserably would mean them not using enough time, which we don’t find to be significant in predicting winning TT.

    Bonus:

    How to name your baby so that it has a higher chances of winning TT?

    
    Regression results
    =========================================================
                                        TT                   
                           logistic             probit       
                              (1)                 (2)        
    ---------------------------------------------------------
    Timur                    0.916               0.512       
                            (1.162)             (0.685)      
    Vadym                    0.852               0.474       
                            (0.528)             (0.308)      
    Nataly                  0.916*              0.512*       
                            (0.489)             (0.286)      
    Alex                   1.322***            0.756***      
                            (0.408)             (0.246)      
    Andrew                 1.253***             0.714**      
                            (0.476)             (0.287)      
    Constant               -2.015***           -1.187***     
                            (0.129)             (0.068)      
    Observations              679                 679        
    Log Likelihood         -269.486            -269.486      
    Akaike Inf. Crit.       550.971             550.971      
    ---------------------------------------------------------
    Notes:            ***Significant at the 1 percent level. 
                      **Significant at the 5 percent level.  
                      *Significant at the 10 percent level.  

    It is true that not everyone can win and one must use every trick to their advantage. And if your name is Alex or Andrew you should use it to your advantage.

    If your name is not Alex or Andrew, than maybe it is not for you. Go apply yourself in some other domain, for you are against the odds when it comes to winning TT session.6

  • 6 FYI, this is a joke and not meant to be taken seriously.

  • Conclusions

    • Time does affect whether your speech is successful or not. In situations where you have limited time, you better take as much time as possible, but not a second more!

    • You can’t cheat the system by choosing to speak in any specific order. On average, you’re just as likely to win speaking first as you are speaking last.

    • There are no differences in speaking patterns between men and women - they speak on average the same amount.

    I do hope that you, the keen reader, have found this piece entertaining and useful. Be mindful of time, and learn to use it to its fullest extent—and not a second more. Good things will come your way!

    Before you go…

    If you like this project and feel like you can contribute with data and/or advise, please send me an EMAIL 😊

    Citation

    BibTeX citation:
    @online{yudenko2023,
      author = {Yudenko, Vadym},
      title = {All the Time in the World...},
      date = {2023-08-04},
      url = {https://vyudenko.blog/posts/toastmasters-time},
      langid = {en}
    }
    
    For attribution, please cite this work as:
    Yudenko, Vadym. 2023. “All the Time in the World...” August 4, 2023. https://vyudenko.blog/posts/toastmasters-time.