I somewhat agree with what candelario_e is saying and what others have said, but I thought I would add my $0.05 (since they just eliminated pennies here in Canada) from an Electronics Engineering Technologist *and* Computer Science perspective:
1. There is a physical link spec, that is where the different cabling specs come in. As we got faster physical link speeds, the signalling rate on the wires had to increase as a result, hence the need for better specifications for the cables themselves. (Note I did not say BAUD rate, as there likely are encoding schemes involved, just like in RF.) The number of conductors and so on are the same, the RJ-45 connectors are the same, but the impedance, twists/length and the requirements at the crimp end etc. have changed. The old wires will plug in and work for short lengths, but you may have lost or corrupted signals meaning lowered throughput from retries. The wire could be a limiting factor in terms of attenuation (signal loss) over distance, but typically you hit walls in other areas first. This is where the original repeaters were used, but they only boosted the signal, they didn't have intelligence and buffer and re-create the signal per se. This is also where using fibre helps, light travels much faster and as a bonus is immune to RFI / EMI (and a lot of environmental concerns like water, humidity, temperature and so on). This plays into what I discuss in #2. As an aside, more rubber may attenuate interference a slight bit, but it is more the shielding and proper twists that helps with immunity to interference in a twisted pair cable.
2. There is a signalling specification which defines timing and maximum "windows" in which signals must be received to be considered valid. Where candelario_e was saying is that the electrons travel at the same speed, no matter what the cable length. If you put too long of a segment in, you violate the timing constraints of a valid Ethernet conversation (CSMA-CD) even though only one unit talks at a time (except possibly in full duplex mode). The frames get lost or corrupted and retransmissions and possibly even collisions increase because of these timing problems. (Station B is listening because it wants to transmit, all is silent, so it begins transmitting, but just as it does, the signal from Station A arrives because the cable was too long, oops, collision, set the back off timers and start again.) There comes a point where it is unworkable and virtually no data gets through. Ethernet does *not* have all the same numbering and sequencing and retries that a higher level protocol has. It only tries a frame a small number of times before giving up. Now if you use fibre, since light travels faster than an electromagnetic signal in a wire (about 40-70% of the speed of light in twisted pair), even with the same signalling timing, you can go further.
3. #1 and #2 are the Ethernet levels, there are the protocols on top of that, nowadays typically TCP/IP. It has checks and retries and all that built in for reliability, however having to do all of that because of problems at the lower level reduces the *real* data throughput and makes the connection "slow".
4. The advent of switches changed the whole "maximum number of repeaters" and "maximum length" scenario. A switch takes the signal in as an end-device (much like a PC or router), it then starts the sending process all over again on each segment it intends on sending it down. Therefore each connection is subject to these maximums. The only limit to length becomes the timing of the overlying protocol. At past companies, we did a similar thing and wrapped TCP/IP packets into our own protocol with huge timing windows to move data over the satellite network. Each end thought they were talking to each other, but in reality, they were talking to our devices, which talked our own protocol in-between them. Of course, if you put something more intelligent than a switch (or a Level 3 switch) in-between, then they do the same thing for TCP/IP (or whatever higher level protocol) as switches do for Ethernet, which is how we can connect the Internet around the world and beyond. (Think International Space Station.... yes, they do data and VOIP.) This is also how those "magic" long-haul ethernet boxes work. They use their own proprietary signalling and protocol between their boxes which was designed to work over the longer distances.
I hope I haven't really confused anyone like I often do when I try to expose the underlying truths. Oh and by the way, the CompTIA people aren't overly technical, their business is to write the exams and books, just like what you read on the Internet, always consider the source of your information before putting your reputation on the line based on it. Check out the specifications from EIA/TIA and IEEE and the other organisations who are behind the technologies and trust that way before a "Readers Digest" version that wants to give you the easy answers.
Oh and where candelario_e is wrong, many devices can talk simultaneously in a multi-port switch with a true switching fabric, as each port acts like a separate device with its' own buffer etc. That is why the switch fabric is rated for slightly over the combined speed of all of the ports in a decently engineered switch.